Quantum Conversations By the Bay II: Simulating Quantum State and Computations with Kubernetes
so it's really the role is about and the team is about making our systems and our technology more available for researchers and the research community as well as the academic community for education and I'll give you some more insight into how we do that when when we went for done introductions awesome next person is Ali listen to yourself Ali you're a mute Ali are you are on mute yes I am fine and you know I can't say much thank you I'm Andrew miserable hi everybody I worked for Chase and they work for Council I know a little bit about quantum not hold on I'm just going to no thanks awesome thank you Andrew T Andrew T it's it's me hello I'm glad to to hear everybody it's nice to meet you I'm business development manager at one to machine learning project that's Russian quantum center and I'm here just for listening and new knowledge ascend to to meet everybody first meet did you hear me hello yeah it looks like Alexi might be like we heard you definitely so Lexi may be having issues again so let's do jacob Mendel please good morning everybody good evening if you are from the other side of the world I'm from JP Morgan welcome thank you our own our own uncle yep not our enough yes one two want me to introduce myself please okay so I'm hello everybody I'm a researcher at German synchrotron which is every region is des why we call it Daisy it's in Hamburg Germany and I do research on mostly quantum dynamics and how to do multi-dimensional spectroscopy with entangled photons so I'm a theoretician so this is the branch of my research and I'm pleased to join the event welcome welcome vetali vetali hi my name is Vitaly Durham I work for JPMorgan and work for Constantine actually so I am a software engineer I was previously software-defined networks engineer and that's my spiel in terms of what I do the rest is gonna be via demo I'm from JPMorgan Chase I work from Marco and I'm software engineer currently but I work on quantum algorithms great Constantine yeah I'm with JPMorgan Chase I did that thing called advanced computing that among other things includes quantum computing since two years ago and we also do you know cloud machine learning and all of the rest of the advanced stuff awesome thank you awesome tea Gilliam yeah hi I'm Austin I'm a software developer and a researcher with a background in machine learning and now quantum computing as of about two two and a half years ago and I work with Constantine in Batali and get to know about quantum and probably make it my career thank you welcome in grenouille hello I work in JP Morgan Chase a software developer and recently I got introduced to quantum computing through one of the sessions so I Marco and I am very happy to be involved here and listen to you on this field which I'm currently exploring thank you you may be lost Alexei again let's see how about Greg hey guys so I work in the 80s of big data applications intelligence and machine learning I'm interested in applying quantum computing term probably maybe distributed systems and I also want to write maybe a condom and grantham at some point of time yeah Jacobo eversince hello I'm a member of the machine learning center of excellence in London working in JP Morgan or Chuck Wang I'm looking forward to learning more of quantum computing in applications to machine learning thank you welcome so as you guys noticed the some four connections are a stable if I'm opposing a Sebastian will take over deleting the introduction so don't worry Jamie Jill what hey let's see I'm a Jameson students okay how about Hagen Cathy hi I work for JPMorgan Chase and I work on engagements for the future lab of Applied Research and engineering with Marco pistoia and I have very minimal knowledge on quantum so just here to learn and listen thank you Thank You Marco pistoia hi everyone I work for JPMorgan Chase I lead the future lab for applied the research and engineering also known as flair you guys have heard a lot of people from my team here and I lead the quantum computing effort the research effort at JP Morgan thank you great to see you Marco Michael Maynard hi I work for JP Morgan and I'm just here for informational purposes newly exposed paradigm so I'll just okay it looks like Alexi my friends yet how about Nick Braun hi I'm Nick Braun I been an experiment of quantum computing IBM since 2013 I currently work on Hardware engagements with startups academic groups companies and education training great Thank You Rajesh narasimham hi everyone I'm a software engineer from JPMorgan Chase and I work with Constantine and Austin primarily on cloud and machine learning technologies here to just learn how quantum computing is combined with cloud technologies thank you welcome Ricardo Carrillo hey guys we got a career here I'm a junior engineer working with JP Morgan Chase and the quantum and ml team in digital advanced computing decided to take me another way so I'm here to absorb as much information as possible just to try to be a useful and then what they're doing awesome welcome I think we're all learning so this is really great to have you know so many folks learning this field that's amazing Sahana oh all right so again I work for JP Morgan and Chase as a Java developer with some I'm a little background quantum is something that is just started exploring welcome Susie Sam hi Susie you're a mute I think oh sorry yeah so I'm that I'm a Java developer I'm just listening in try to get some information awesome welcome to do no funky hi yeah I'm here merely for my curiosity to learning I'm a data architect with JP Morgan but I'm also studying school at Johns Hopkins in end of August for Applied Physics so wanting to learn more about quantum physics and quantum computing so great welcome let's see Simon doing all right my name's Sam Ewing I'm a graduate level theoretical quantum chemist so I'm working in an area called reduced density matrix theory so it's only tangentially related to actual quantum computing but I'm excited to learn more awesome welcome Joachim score it's hi I'm also at JP Morgan Chase working in model development and interested in quantum computing entirely out of my own interest thank you sounds great and you have a lot of interest receiver piano in the frame which is Vasily Nechayev silly yeah nice how much 20 bases finished on the ground just about quantum computing and its application great welcome we need to talk when it's all high so I'm an employee of JPMorgan Chase I'm into architecture and design authority space but I'm interested in this quantum computing going through Constantine and Austin papers as well as Marco so you have videos and papers so yeah I am pretty much interested in this piece and trying to get information out great welcome vgg I think vessel in your give yes hi Alex my name is Vaseline and I just entered the channel so you kind of quickly notice that I arrived what are we doing we've lost Lexi Sam just a quick intro just we're all just introducing ourselves okay so my name is whistling your gif I'm located in California more precisely central California my background is physics and I've done nuclear physics and computer modeling and recently I'm interested in quantum computing because this is a great tool for me to continue my explorations of quantum physics and nature welcome you son hi everyone I'm with the Condors research and racing Morgan Chase and working and working on quantum computing with marco constantly and opting to have to be here right well it's a young saeng sorry I just saw bill to myself hi everybody my name is Geum JA and work and a Stanford researcher computer center and my background is in high-performance computing and computational chemistry and to recently have been working on a new quantum computing algorithm for or quantum chemistry thank you for this great opportunity for learning more about beauty thank you I'm not sure I caught the audio but it's maybe my connection we heard it fine told you guys it's yeah you know know right this is the uncertainty of our online times about to have a lot of good folks here who can ascertain everything well so I think at this point if I miss anybody please introduce yourself feel free to we open the floor for the rest of the introductions feel free to speak let's see if that's this will work or you can also mention it in the chat and I can read your name whatever works for you so just let us wait them a moment or anybody we missed just type your name in the chat and I can introduce you if not I think we will proceed to the next item so we're very fortunate to have with us Sebastian huttinger who is the lead of academic programs at ABM I and basically he thought of the squander converstations was his idea we did a lot of things together and previous like this and he will say a few words about the about his work sure thanks Alexi I know I wouldn't say it's my idea necessarily I feel like this came out of conversation with you so I run the academic partner program for IBM quantum there's a lot of I mean it's very simple to sum up what our mission is but it's also very complex to explain all the ways in which we're doing that very simply put we're trying to expand the community around the IBM quantum platform and technologies to include education researchers industry and you know increasing the general level of awareness and knowledge about quantum computing now translate that into programs and offerings is a couple of slides worth of tons of links so you know I would say it's interesting actually this workshop this particular session it's great to have so many parties from JPM see this this puts more of a an industry partner slant on it but we start sort of with learning resources and educational resources we've got a tremendous resource in the open service textbook and the YouTube channel as well as self-paced tutorials and stuff on the IBM quantum experience so whether you're you know self-paced learning or you're you're looking to set up a class in a university or actually in industry as well we've got all sorts of resources that can be used to to support those activities we're also developing actual offerings ourselves we have a group called the Skills Academy they have a first quantum computing offering which is sort of a turnkey solution for delivering sort of an introductory cross-disciplinary view of quantum computing so it's it's a starting point for institutions and and and firms that don't have existing capability we'll be adding more to the their offerings so that there is we actually are working on a quantum developer certification test and we'll be building almost a preparatory class that will be able to to be geared towards that that certification test and it's all coming up in 2021 so there's a ton of activity around around education the the other sort of focus is around research which is probably more cogent to this particular audience so obviously there's institutional access through the Q Network membership JP MC is a member of the Q Network and that that gains you access at a at an institutional level site-wide level to all of the private resources that that iBM has the 53 qubit machines of 27 cubed machines open pulse dedicated reservations as well as other other support and and content but then that's kind of the commercial offering what my team is doing is try to create programs that get access to individual resources at the researcher level so if you're a researcher that's affiliated with the institution that's not part of the Q Network or you're not part of a project that's at Oak Ridge for example or other ways to gain access to the Q network we're trying to make it easier for researchers to be able to gain access to open pulse on a larger qubit machine and in some instances the larger machines we're needed for for supporting a particular project so in July will be actually launching a couple of webpages that will take applications for membership to the IBM Q experience for researchers group which is a higher priority Q on the public machines as well as machines that are not available to the public and also a private slack channel for sort of talking amongst ourselves about how to how to use these systems and also access for projects which if a project requires open pulse or pulse level control whether you're looking at gate design or error correction or simulations or other applications of the the microwave pulse control we can grant you access on a project basis to a five qubit machine using open pulse and if you haven't checked it out open policy is available to the public but it's on a single qubit machine so you can get a sense of what it you can do but it's limited in the applications obviously because it's just a single qubit and then in addition actually I on the education front the third program we're launching is if you're a professor that's using the open source textbook we can grant you access to a group that has access to that five qubit open pulse machine and that you can provision your own students on to that group and also scheduled dedicated time on that system so that if there's a coursework or classwork that requires faster turnaround on the Q you can actually line up at that dedicated access with your class time and in addition there are other programs we we run like the PhD fellowships which are coming up in mid-september and and also faculty grants for being able to award financial and in-kind services for projects that we support so there's a as I said a large number of ways to get engaged the overall as I said the overall mission is really just to expand this community I think that from my perspective is sort of it to close out my comments I think that we're in this really fascinating time right now where you've got basic fundamental research going on as well as the early stages of commercialization and the attempt to build the infrastructure that we need in order to develop the workforce that we're going to need as as as industry clients as IT vendors and as academic institution so there's there's a number of challenges right now that really iBM is view is that the best way to tackle those are is as a community and as partners and is with as much open collaboration as possible so that's really our approach to building the communion and that's the the essence of this workshop is really this start having a regular check-in that people can you know start to develop connections get exposed to new ideas new work and in a more informal way back to the the starting that the conversation with Alexi what I really liked about what Alexia had done was that he brought the data science and AI community together in the Bay Area in a way that was highly collaborative and in a less formal way that I think is can be a little bit less familiar to the the the hard science research community and I think there's a lot to be gained by sort of cross pollinating those industry players with the research with the IT vendors in a more interactive fashion so we can really start to get to know each other what what work we're doing and and find ways to collaborate so welcome Thank You Sebastian I really appreciate it and yes I think you know this idea was really resolved the collaborative brainstorming we did in the coffee shop we remember there were coffee shops in the previous life and so in the Blue Bottle and originally we wanted to make it a physical community a quantum West right because the East Coast is heavily represented but the West Coast is not so much in the community space right and around the by the bay conferences and meetups as Sebastian said and we really have great image experience so most people come back and say you know this like 30 40 percent of the people in the conference stay in the hallway so there is a famous hallway track they meet each other and establish this lateral connections and I found when I go to academic conferences this is often missing because it's a bit hierarchical it's a top-down and it's it's it's a great culture but you know we want to kind of spruce it up and introduce computer science influence and it wishes you know this kind of bottom-up graph of connections which builds by collaboration so yeah and Alexei just before you turn it over to JP MC to present you may want to give somebody else co-hosts because I'm gonna have to unfortunately drop at the top of the hour so a little over 30 minutes sounds good yes so I'm gonna probably yes I'm gonna I mean what I'm gonna do and probably just give it to Constantine and he can hand it off I think Austin would be a good choice Austin okay so is our backup in case something happens great great yeah you guys come prepared I really appreciate it and so Sebastian one more question for you if folks want to contact you about the program what are the best ways to contact you and I can share it later with the list yeah just email me is fine I mean I can share that slide with the links so you can explore those links actually you can share that to the group Alexia after all those links are live so if you want to explore what's available that's a good starting point if you any questions just email me grace oh so basically I just you know send me the slides guys and I have the list of all folks register several forward the slides but if the slides will be available and if Constantine will make their slides available I can share that as well right so basically anything speakers care to share we'll share later through the mail list and obviously co-op Thomas V is our website we mirror information there so this is where the videos appear after the meetings and so we'll put the video there as well and if you want if you're looking for something and don't know where it is just email me my email is there is Alexia chief scientist at work and so I'll try to find for you thank you okay so now we are going to here our main talk it's by the JP Morgan Chase Group and Constantine is a speaker we know well in the sky leukemia so I think it's really really great to have someone who struggles so many different worlds so he did the talk on basic electronic material skull and I remember vividly when it appeared it was the most unlikely combination of all things you can think about but you know on the other hand I think it's a likely combination because a Scala is has this mathematical quality to it which appeals to a lot of scientists so for instance a lot of national inch processing is done in Scala because the kind of folks who do NOP tend to choose calipers conciseness I think it's kind of it's almost like mathematics so I found that a lot of folks who were trained as physicists and mathematicians gravitate towards it so I think it's really great that you know it found its way into quantum and then broader context function programming and so to me this is an amazing achievement that Constantine is able to do things like Scala and quantum in a bank and it's really interesting how this all applies to fintax so welcome Constantine and take it away so you can you can see my screen yes okay so I should say that we are partners with IBM IBM actually helped us get started two years ago in quantum we still have that partnership and we meet regularly and so on we have papers you know published in in collaboration so that's always great to see I be and people here so this is about simulator side we asked Alex do we want a quantum purely quantum talk or something that's combined with computing and he chose this one right so why simulators when you run a computation a real quantum computer that's pretty expensive right so one reason is you need to prepare and make sure that you have your everything is verified it's correct and so on but we found also that simulators are a good research tool right so you can just poke to to the quantum state and find patterns or some insights that otherwise you wouldn't see them and that's why you know I use Scala as well we actually had a Java simulator of course we have Python and one thing we found out it's a very good tool for research and we actually create simulator specific problem now right so we will talk about that a little bit and in this case because we were looking into distributing and having as many qubits as possible we chose girl and kubernetes but writing you know a simulator as you'll see is not very difficult so first of all quantum computing is very good at solving exponential type problems right that classically are very hard and I'll just use a simple model or easing if you are German so that the the model is basically you have a number of sites that you can think of them as magnets and then magnets can be it can have the North up or down right and this is an exponential number of configurations in terms of number of sites because each magnet can be up or down right and then if you want to investigate these configurations then it's a very hard problem and Cuong quantum computing is very useful in this kind of context and for such a problem you define a Hamiltonian which assigns to each configuration and energy it's based on the weight of each interaction between magnets and you can add this gives you a quadratic part then you can add interaction with the environment and that's a linear part so throughout the presentation I'll use this simple example and the interactions will be only between neighbors you know magnets that are that have a North Pole up so basically is the number of neighboring magnets for sites with a North Pole up right and we so we are interested in investigating ground States which means was the minimum energy configuration in this case it's no neighbors should have you know they're not Korra because then you get zero right so those are the ground state how many are they and how do you get them all right and and so on so this is just a very simple example I'll pause from time to time to ask for questions but if you if you feel like I should clarify something please let me know what is quantum state in general and how do we represent it in a you know program in a simulator so we are not trying to address quantum mechanics here quantum computing is simpler than quantum mechanics all right so for example in quantum computer you only deal with binary strings as outcomes and basically it's called the computational basis where the the basis is form of the binary strings of a certain length right so this obviously matches you know the view of computer science where everything is based on zeros and ones and that's that's where quantum computing is you know Canaries connected with classical computing we're not interested in other kinds of computational other kinds of basis or other kinds of quantum state or even we are not interested in for the purpose of this talk how it's implemented in a real quantum computer right so with that being said the quantum state is very simple it consists of pairs of outcomes and complex numbers called amplitudes for each of the outcomes right so for example if my space is 3-dimensional I have three qubits then I have eight possible combinations of zeros and ones and to each of those we associate a complex number called amplitude and that's it this is the quantum state will the constraint that the squared magnitudes of those complex numbers they add up to one right and with this that all we need to say about quantum state from the computing quantum computing point of view pretty simple this is a spreadsheet basically right so another useful thing we found out when we play with you know quantum state and quantum computation is we need to visualize what's going on right and we came up with this idea of associating a color to a complex number so basically a pixel if you want the hue of the color is the angle the phase of the complex number and the saturation is the magnitude right so the more intense the the longer the absolute value of the complex number the more intense the color and of course this is not new complex analysis is using this for some time right but the application in quantum computing we were surprised to see that nobody else did this and actually were contributing this into kiss kit which is the IBM framework and iBM has a few types of visualization this the next illustrate Austin you you are working on that that's what worked hard that's me I release I think iBM has color histograms what this was our first idea why don't we color the histograms right so it's already there now in terms of colors we start the real numbers are red because we start at the x-axis and then you rotate right I think IBM chose blue for the histograms because the normal color is blue and then or maybe because its idea and then you know it rotates a certain way so colors are a code basically and since we are talking about outcomes when you have a quantum computation I guess most people know this but I'll just mention it for the people who may not you get when you measure you you measure the quantum state the quantum state is collapsing to one of the outcomes so you'll just see that our comes a bunch of zeros and ones a binary string and then that quantum state is basically destroyed it's a classical state so you you can create it again you run it and you get another outcome right and then another and another and another based on the the length of that amplitude you know the probability is the square the square of the magnitude and this is called the born rule it has a pure intuition and part of bonds you know when she created this initially he thought it was just a magnitude but then he found out its squared right so it's it's very similar to a slot machine if you want so when you pull the lever you you create a computation and then you measure it and you get one of you know the binary strings and then you do it again and again and again right and if you build another image with with the pixels that you got you know because each of these can be mapped you there are the keys of the pixels then you get a sort of gray image of the intensity right that's how you can think about it and it's useful since this is sort of a strip you know a column of pixels we we like to dimension in generally it is easier for us right you can make it n dimensional if you want any that's the real quantum state if you want it's n dimensional but the best for visualizing it is if you make it an image so you you have two dimensions so you basically split your your amplitudes into columns and to make it nice we usually make it prefix so each column has a prefix and then the other side the suffix of the binary string so this way the quantum state really becomes an image right and it's not an analogy it's really want one you can think of you can convert a quantum state to an image and if you work hard enough you can convert any image in a quantum state but you have to normalize it so intensity is add up to you know the squares to 1 how do we represent this in code so you go and it's essentially an associative array right like Java map or Python dictionary all of those who work what we found out in in go if we use the go map you know the built-in it has concurrency built into it so when you add the key remove a key to complain right here we don't you don't need to add keys or remove keys the keys are always the you know the binary strings of length n 2 to the N right all right so then we are not we don't have concerns about concurrency and a binary tree works great so that's what we did basically took the implementation from programming 101 you know you never know then the tree and that works great so if you take binary string you follow the path like 0 1 zero one zero right to follow that path and you give you amplitude which here is zero point zero two plus zero point zero nine times I write the value that it believes the amplitudes what about any questions so far about the representation basically a map or dictionary okay so what about a cube it is talked about a qubit in a you know in quantum computing how do we represent that well we don't need to because it's just a position right in the binary string there is no entity or something like that so in physics you have an entity right a particle and you act on it you write but that that gives us a dimension that's how we have to think of it in quantum computing it's it's a dimension so in math you know how to apply you know operations across the dimension and that's why the qubit gives you is that position so there is no special representation for a single dimensional one dimensional system or a single qubit system it's just to two pixels or two complex numbers right associate to zero and one and that's it this is a you know a qubit is a dimension and what are the single qubit gates or the one dimensional gates you are probably you've heard of Hadamard or XYZ and so on right gate is just a function right that takes two complex numbers and replaces them with other two complex numbers that's it all right so those two pixels you trade some intensities and colors and you get two new pixels that will replace the previous ones and that's all there is to a gate from the computational point of view right and the basic basis gates in general like in any instruction set in any kind of computing like you know a risk and so on you have a set of basic instructions and the rest are derived in quantum computing you you can work with H atom are than the phase gate and everything else is derived and the implementation here in go you see how the Hadamard is basically you have two complex numbers a and B you replace them by a plus B and they - be normalized because they need to you know the squares need to adapt one so you replace two numbers by their sum and difference and that's Haram at the phase gate rotates the second and one so you multiply the second complex number by another complex number which means that you rotate and that is a unit complex number so you multiply by cosine theta plus I times sine theta and that's phase the phase gate and here I can run an example so the initial state we usually we we start with with the state that always gives you zero zero zero in this case I have a single qubit so it's gives me zero that's why you see the highest intensity and it's red because it's a real number so we always get zero in the initial state it's basically not a quantum state if you apply Hadamard you'll get a superposition where both 0 and 1 are possible alright and the probability is the same that's why the colors are the same it's a little bit less intense than the previous one by the probability is 0.7 and so on is 1 over square root of square root 2 and then if you apply a phase gate to this one you'll get you'll see the second number changes in color so that means you multiply it by a number and I can write I can run something in the ID any questions of our so I can run this let's look at the code so I want to have just one one qubit one dimension and then I'm applying the H gate to well it's only zero yeah I can have multiple qubits right and then we'll get an image here if I ran it and I'm saving to an image and that image is here right and you can see we got the same let's see if I if I use two qubits yeah you see so now it's a it has four pixels and I can actually make it two columns here if I if I want I can it has to be a power of two but otherwise you have freedom and now it's it's a different shape right so this is what the simulator does for H you can you can play with it understand it better right but now let's take the face so the phase I have another test for phase right I'm doing Adam are first because otherwise if the second qubit is not active I I won't get anything right so I don't have them are first and then let's say three PI over eight well if I run this I'll get this this picture right let's change the let's make it 3 PI over 4 look at the image did they say okay I have a mirror I didn't take the change okay let me try just PI over four yeah and you see a different color so that's how that's how the phase works and I can you know I can use multiple bits right so it's already you know giving you an idea about what what's happening any questions so far so the next thing is the other gates the other one dimensional gates how do we implement them right so if you have h and the phase you can implement the rest i BM actually uses instead of 82 these are X PI over 2 which is a similar to H so it still takes the sum and difference but it rotates them here it's a multiplication by I in each you know each of the amplitudes and the other systems they for example honey values is different or trapped ion right they use different bases in Rx and ry combination and RZ right so the beauty of a simulator is if you work with a different with different hard way then you can adapt you can put new gates in right it's as easy as adding you saw the go code there and after that can build rxry yeah I have you know my notes here but they are not hard to derive for example X X is just swapping the two amplitudes that you have a and B and then you return you have to swap them and the Z you just change the you know sign in front of the second amplitude right and with a simulator you have a choice you can you can do this and you have a shortcut or you can build them from the basis gates and then you are more like a transpiler right again you have a depending on the goal of the simulation you can take some shortcuts if you need to what about multi qubit gates right so how do we transform quantum state in general and the answer is you don't need any two to be n by 2 to be n matrices so that is matrices are useful and linear algebra when you when you prove something or you read proofs right but you cannot apply them to a real computer the only thing you can apply it to a real quantum computer is a single qubit gate all right that's it so then the simulator should do the same right and of course you have control gates we'll talk about those the IBM has a third which is the C naught so how do we apply a single qubit way to a multi qubit system right and the answer is that you pair the amplitudes so you choose a target position and that's your qubit right let's say the middle one here if I have a 3 3 cubed system or a 3 dimensional system the outcomes are binary strings of length 3 so if my target is the middle then I pair up the outcomes in such a way that they differ only in the middle position which is the target and the rest are the same right 0 0 1 0 1 1 and then I applied the gate formula whatever it is X Y Z to those pairs writes the single qubit formula I apply it to these pairs right so this change is all the amplitudes right it's a huge amount of computation right if you have billions of amplitudes so a single single gate application in an n-dimensional system changes everything right and this is the power of quantum computing physically you apply it to a single particle right but computationally you change the whole you know billions of numbers and in classically this is very hard so that's why the simulator will will take a lot of resources and the converse is true so you know in a quantum system it's very hard to change a single amplitude very hard right but classically we just go to the record and change it in a database right so this is the main difference between classical and you know quantum computing quantum computing is multi-dimensional so what we can say you know by these pairings what we what we do is actually we just have a for loop we go through all the amplitudes you know let's say we look at those with zero in the target we find a pair and then we just apply the formula and that's it this is the only computation you do but you do it you know in a giant for loop so quantum state transformation a unitary if you want is just follow you can I mean you can unfold matrix multiplication into four loops in my talk you know in Scala that's what the you know that's another way to look at it monad a monadic approach right any questions so far so now what control transformation in a physical system this is entanglement right you it's very difficult to do it but in a simulator what you do is actually you restrict the application of a gate to those outcomes that have some something fixed in some position like zero and one you say first position has to be zero third one one and I only apply my transformations to those classically this simplifies the the computation right so it's a shortcut that you want to take in a simulator in a you know in a quantum system on a real computer this is very difficult so the number of controls actually kills it's a high number kills the coherence but in a simulator this actually helps right so again depending on your goal when you simulate you can take some shortcuts and this is one of them if you want you can you know have Angela's and some but you don't need to you you can even match zeros right no not only once so iBM has an additional gate called C naught or control X to do this so in the tree interpretation of the implementation we have this in code so everything everything I'm saying is in code right and it's you can implement a simulator in a hundred lines of code and you'll get the same results you get on a real computer right so these simple concepts you know pay off and also help understand you know quantum state in general maybe even in nature so if you control on the middle qubit and you say let's say you want one there in the graph in the binary tree you only look at the paths that go through one right so you only those states will be affected there's the interpretation now terms of circuits I have an example here so a circuit you know is represented by a wire for each qubit and then you put gates single qubit gates and those wires and they also can be controlled that's how we represent what I have here is we apply to the bottom right we apply a rotation around the y-axis of the angle PI over 2 to all the qubits first this will put them in superposition so now each of them will may be measured as 0 and 1 and it's also equal superposition so so then what this circuit does you know going back to the Ising model I said we don't want neighbors to be 1 so if Y if the first qubit is 1 I'm I'm doing the superposition so I'm rotating back by negative 90 degrees so this way the second qubit will will measure 0 if the first one was 1 so I'm ensuring that you don't have neighbors you know better one pairs of consecutive ones so we do this on our qubits and so this is an implementation of you know what I mentioned that Ising model simple Ising model how do we represent this and I'll run I'll run this circuit how do we represent this well a transformation has to have a target has to have the gate and also a number of controls I'm allowing 0 and 1 to say in these positions you want 0 or 1 right and that's it after that the state actually I made it an alias of 3 and adding these methods where you apply a gate and each the apply methods are just for loops where you pair up amplitudes and creating new amplitudes before running the circuit make another connection which is more than a connection so any quantum state is an image if you visualize it like this and in go by the way I'm using a pixel really pixels so this becomes an image using the go language that represents the quantum state so every quantum can't any quantum state is an image but if you evolve the quantum state through gates you get multiple frames if you want so quantum computation is actually you can think of it as an animation again is not it's not an analogy it really is if you if you put together the frames you get an animation and now anything you know about computer animation applies to similar to quantum simulators right because you can distribute state so in a computer animation all the pixels change right the same is true it's a huge amount of computation you also have strict rules about what you do like ray tracing or whatever right here we have gates but I would argue that quantum computation is much simpler it's probably the simplest kind of computation where you just you know change two numbers and there's a reference a fork a frame corresponds to 23 cubits a 5k to 24 so it's just one more qubit this gives you again the idea of the power you add more one more two-bit you double the resources that you need in terms of memory you know network everything doubles with every single qubit right so we'll be able to simulate you know to some extent up to twenty-something qubits on a single machine but then if you want to scale that's why we're using kubernetes right so you can add 10 cubits with with 1000 container 1024 but if you want to add 20 cubits you need a million containers right it's not easy but it's doable if you really want it what what we did first thing go was just paralyzed using go routines you know and that again pays off depending on the number of of course you have so you assign subtree to each each go routine right and do you really get a speed-up so that was the first step then we partition the state and then so we let's say columns so each prefix corresponds to a column and then that column is managed completely by a process in terms of memory storage if you want to do storage this later will become containers right and the nice thing about this prefix suffix partition is that when you when your target for a gate is a suffix everything stays in that column so everything is local then if the target is a prefix so again you have to match pairs that differ only in that target it means that the prefixes will differ only in one position so that means that you are pairing up two servers or to two columns right and all your pairs will be between those two so it's a perfectly symmetric computation right if you so first of all you have to if you can stay in one container so if your targets are suffixes right and you can reorder your qubits let's say if you if you know for sure that 20 qubits are not targets they are Ansel Azure then we make them prefixes right and you'll only have local computations you know not even have you know distributive computing so that's one of the optimizations so Vitali will show how will do this in containers and going from a simple state to a prefix state or a column is very simple we wrap the state but we need to find the other container or column that has the pair so we need this amplitude store resolver we called it so let me run this circuit now right you know here we are actually revealing that that Ising model is related to Fibonacci numbers so if you look at how many binary strings have no consecutive ones that is a Fibonacci number and it's a you know interview questions sometimes right so this circuit actually will only produce binary strings you know consecutive ones and if you count it to be human for example here for three cubits we have eight pixels right and you can see three of them are white that means there they are not possible the other ones are possible because they they have no consecutive ones right so I'm showing how we run in one process or two processes so let me do this in the IDE and I guess we'll have to speed this up a little bit how are you doing at the time Aleksey we're fine right I think we're flexible so this is the it's called the test it's called test heuristics so it's heuristic because you you saw how we came up with that circuit it was more like an insight let's put it in super position then let's cancel the super position right it's not canonical or you know automated way to do it and that's one of the you know beauties of quantum computing and what's more difficult even though what I'm showing here is that the state is is pretty simple but what gates to apply that's the really you know hard part so if I run this and I look at some images that were created you will see that there is a single image here and then I if I run the distributed it will create two images right and the columns are the same let me change some values just so we see that so if I make it now four cubits I should see a different image yeah so with two columns and here if you count you so have 6 plus 2 8 right so f f of 4 is 8 right out of 16 binary strings 8 have no consecutive ones and the next one should be 13 so let's test that if I make this 5 and I can make it more columns so we can count I can even make it 8 columns yeah so if you count 3 so we have 5 times 2 plus 3 which is 13 all right and we are writing the circuit itself you know it's just straight that apply you know when you put the gate it's similar to kiss get you know kiss kit applies gates to a circuit right we work directly with the state here and really you can have a dsl like in scala we showed that the SL that is nice or a functional yourself okay and the distributed on it's here on the slide I won't run it but there is a second matter I mentioned that the first was the first one was heuristics so we had to come up with you know good idea about the circuit but if you want to automate this is our latest paper this was the list last Thursday you can see here Austin Marko and myself and basically we said if you have a formula for your Hamiltonian then we can encode it efficiently right so what is the Hamiltonian for this that IC model if I want to count so I map each configuration to the number of consecutive neighbors you know with the North Pole up or in binary strings consecutive pairs of one right and you can represent if the sum of X I times X I plus 1 this product is is nonzero only when both are 1 so you get basically the count right if one of them is zero you get zero so now I have a formula and I won't go into details you know you can read the paper about how it's implemented but when we implement it we we get this picture that you see here when you visualize so each column basically is I should say prefix is it's like the input this is like the graph of a function you have the input and the output on the vertical but it's you know flipped in the paper we actually did post processing so it really looks like a graph here so numbers go up the positive was but you if you count you see that we have 5 with 0 pairs to 0 1 1 and 1 1 0 they have 1 pair of consecutive ones and 1 1 1 has two pairs right and this is all encoded in the quantum state why is this useful once you have this in the quantum state you can find the ones that have 0 0 pairs right which is one level there using quantum counting right or if you want one of them one ground state this is an important problem in the Ising model give me a ground state right and one of those 5 is a ground state so you use the Grover search and that's in our paper you can get one of those ground states right but including it in the quantum state this is useful so the simulator basically helped a lot I mean we I think we couldn't have done it without you know simulating to find all the you know the right things to do for that algorithm so I'll run that we we call it test canonical because it's one of the general it is interesting they didn't know what canonical is the Ryoka piece and they thought it was related to the can-can and law you know church or something but in math you have canonical forms right so it's standard term okay so now I'm looking for that's right this is what we also have in the paper and actually in the paper if you look here we have a picture running on the real computer you get this gray image right and it's really close let me run a high number just so you see if this works for any okay so I'm in the canonical so I'll make it for now sorry it's already there so now see you guys if you count you'll get yeah whatever so we're looking levels zero which is this also we have one two three four five six seven eight right five you we have eight eight ground states for this Ising model okay any question before vitaliy takes over to show how you make this scale right so many times we are interested in just a few qubits right but if you if you really want if you have a problem where you need to test a lot of qubits then you you'll need you need to distribute the computation just like in animation right you you'll have a farm of servers working on and a problem so there are no questions then vitally you can take over I'll stop sharing okay I'm gonna share my screen and I have my keynote here okay so can I can everybody see yeah I can see okay cool so I'll let you process what can't a skeleton just said and I'm gonna describe how we distribute this computation so Constantine described the domain of quantum simulator and the representation of quantum state using the color implementation is using binary tree and provides an example of distribute distributing the computation using gravity so it is easily mapped to the distributed computation using containers we build on top of the concurrent example and create a distributed version of this simulator in a distributed version instead of Guru team goroutines we will use a containers and we will assign a column in quantum state to a container the name of the container is the prefix for that column right you remember the the grid where you have rows and columns so it will become a direct mapping all as we say in distributed systems it's simply shouting we implemented a handful of APA guys for the simulation we use G RPC for the communication so the screenshot here is a exact representation of the Slims slimmed-down simulator api you have a couple of gates here you have Hallam our phase control phase are why I can control our wine yeah this can be this can be made more generic when you pass a gate like what do we apply but you know for the time being we just kind of hard-coded these gates yeah so two input methods are get on two children set updated right so I know since all keys are fixed and there is no variation variation in the amount of data and you saw the tree representation of it this the main is perfect candidate for distributed scape a gatherer pattern using prefix charting the number of containers is the total number of prefix prefixes or columns so we will delegate the container management in this world to the kubernetes of course and we implemented the custom resource definition using the queue builder that basically gives you a it provides like powerful libraries to and tools to simplify building the kubernetes api right so we have an abstraction which is the we call a server server takes in a [Music] map CRD request let's say which is which is simply a computation that you wanna perform the driver which is the cid in this case we'll create a container topology based on a commutation you provide it and then compute pass store the results and then destroy the containers after your computation so there are a couple of requirements of course we need to apply one gear at a time and we need to acknowledge before passing back the results that all all the gates up lined so with kubernetes you know like it gives us basically a platform for for the container management now state in easy repeatable form right so or you need to manage it manually in terms of retries and stuff like that so yeah so this is the visualization of cube builder now we use kubernetes specific implementation of the amplitude stories over and you saw it in a previous slides with with simple goroutines so instead of the specific go routine we use a container with a prefix name and here you see a random computation and you see the container names with 0 0 0 0 0 1 0 0 1 0 and so on and so forth which is simply live shot now let's do a quick demo so instead of using the CRD i'm gonna demonstrate we split the computation in a couple of parts one using using the simple go client i'll create first I'll create the state so I'll run this command go round prefix demo which which will give me a visual which will which will create containers that will compute some columns right so each container here is representation with a represented with the prefix name and the you will see graphical representation of it right here right so I just created this this state that that is able to accept a some random gates computation whatever I want to send in there let's look at the pods yeah so the same pods are represented here now I'm gonna send a random computation and this is just our X apply I'm applying our X and control our X so and I'll tell the logs so the computation have started you see that it now declined for for for the distributed state send out those gates using those api's that I showed earlier and I'm applying our racks and controller acts now once the computation is complete we actually push the image in a remote store where you can you know like look at it but here in this demo I'll just copy the image locally and show you how it looks so it should show up here yep so for each column constatine show this visual issue virtualization earlier for each column we are creating from each container we are aggregating and collecting the image and this is the full grid image so I forgot how many what is the matrix here let me see so I used prefix of three with the beats four that's why it's a lot of colors yeah and we should say that if you have billions of amplitudes you probably don't want to obligate this on the client you probably want how do you simulate measurement right that's another question now maybe you want the top probabilities you may be the single if you have a million containers you probably want the top in each container and then you create and see which one has the highest probability or something like that right yeah and those are the optimizations yeah well simulators you have to choose what you want to simulate right but you really can customize your simulators to simulate it to the problem that you are trying to solve any questions I guess no questions I have gotten a couple questions in private if you are planning to share this anywhere on github yeah that's a hard hard one for us you know lots of approvals and so on that's why in our first paper we actually put the code in the paper and it became pretty long because that was allowed but so yeah we don't know whatever will continue to kiss kit for example that's easier because it's maintained by IBM you know after that and I guess Marco this is something that we can tackle later on definitely I think this would be like a great tool for for the whole community so we're very happy at JPMorgan Chase like just like a side comment you know I joined the JPMorgan Chase in January of this year and I okay I came from IBM research aspect to 24 years as a researcher IBM Watson Lama so when I joined JPMorgan Chase I was very excited you know like leading the flare initiative working on computing but was not sure you know JP Morgan Chase doesn't come across from the outside as a technology company and it doesn't come across like as a leader in in technology because everybody thinks that JPMorgan Chase is a bank so it was my surprise and I was so happy to find a team with so much knowledge and expertise already quantum computing as you guys can see from these two presentations and we're writing papers together with Constantine an Austen in particular but soon the whole team like us on Alex but Jacob and Al you know there is there are many people that are here in this meeting so you you you it has you know very well known paper and option pricing exactly the investment bank that well what's interesting is that for me this is sort of the third revolution we're going to Mobile was the first one and we actually you know always created intellectual property and we were number one from the beginning we still are for a reason right we we couldn't make it public or talk about it right the same with the JavaScript you know my team created the frame or before react was there right and when when I you know when to to con I think it was talk to the Netflix people and we became convinced we need to do reactive we created you know our own and the same here we have you know a lot of utilities a lot of contributions internally and the difference is now we can also share so we'll see if we can go even yes exactly so of course being a banker JP Morgan Chase has to obey certain regulations so it's always a little difficult to share outside you know every single thing that we're doing but things are changing because the company understands that we are a research community with a research group so we need to be part of the research community and so we're happy to share but I think people now understand how much knowledge and contributions is like are there in the JP Morgan Chase in the field of quantum computing and we're very happy to be part of this community and share with all of you what we have done and learn from you as well all the things that everybody else is doing yeah and also on a personal note you know I was in math I have a PhD in math and then I switch to computers and when I went back to math like 15 years later I don't have the same tolerance for complex stuff complex notation you know superscripts boards and so on and cursive because it's hard to implement I was in code and if the if the goal is to put something code why not start that simplicity and keep the simplicity so that's why you know you didn't see any unit Ares you know or matrix multiplication or linear algebra or you know high dimensional spaces and so on high school high school math that's how our first paper actually had an introduction and the goal was to have high school level introduction to quantum computing and they simulator actually falls into that category where it's so easy to implement it that you know a lot more people should play with simulators and quantum state you know as the organizer I can attest to the fact that you know the moment open-source people generally see sound like lists they have double desire to play with this because first of all it's you know it's old familiar technology right this is a technology which everybody is learning so I think it's an amazing feat because you combine two things you know you made kubernetes a learning mechanism which is just truly awesome I'm so you know I bet if the committee will appreciate I understand how legal ramifications in finance are but you know if there's anything we can help with you know would be happy to for instance now since I'm at the Linux Foundation you know an IBM as a member of the Linux Foundation maybe that's a mechanism right - so one kind of way to basically put open-source under Linux Foundation and that actually makes it much easier and there is a well known way to to you know make sure everybody's happy so just just an idea definitely so I also wanted to say I have a very similar background to Constantine so I also have a PhD in math and four years in years I was like why did I even bother to do all this math like if I was doing computer science major I reckon so quantum computing is a good thing you know for me particular because I was finally able to use the math but I don't agree with Constantine's point of view basically there is no reason for people who are approaching quantum computing from different areas you know you you don't have to have necessarily a PhD you don't have to have a PhD in physics necessarily that's like almost a misconception that pushes people away but for the physical implementation of a computer quantum a you do need right it's the work sure not about color for the work that we are doing for algorithms and applications yeah it helps right but you don't really need even an advanced degree yes exactly so for definitely for the work that we're doing at JP Morgan Chase we're not really building a quantum computer we're not in the field of quantum hardware so having a foundation of physics hacks for sure because you can actually relate to certain concepts and so on but I like personally I really like the do presentations of both Constantine and Vitali because of its simplicity but the simplicity didn't take any rig or away so it's still a rigorous approach without having to go through a difficult path you can further up a constant dimension this is what a quantum state is so the code doesn't like the cloud developers say that my friend of mine at IBM used to say code talks so basically like it's a rigorous approach we didn't take anything because simply didn't take anything away from this scientific procedure but on the other hand the straightforward approach that we have seen in this presentation I think is is in a way even normal and very helpful for a lot of people I think and it's through reactive Alex yeah just to clarify you know reactive foundation is not about just reactive applications is actually we see it as a cloud native applications foundation so CN CF is cognitive computing foundation which is an operator view operating kubernetes as a infrastructure for for developers but you guys actually using it you know as an application mechanism right so we we really want kind of simple api's if you're a developer so you can scale your app without going into all the details this scales a specific use case right a quantum simulator correct correct great so any other questions guys for Constantine and Vitaly and marker no I'd like to ask a question and exit is there any additional advantages assimilating work aid compared to other cluster or canoe or customization and I just make my sauce doc respond I said because of the good integration with say the cloud providers make GCP AWS I mean we I mean GCP and AWS aks are both kubernetes platforms right so we just from the API perspective it's easier to the I mean I don't even think matters is in play I mean it wasn't play maybe 45 years ago but I mean the cube became like a de facto standard for writing distributed application he has a super clean easy API and I mentioned that we use Q builder which is like a framework around the controllers writing pure controller is kind of you know like you need to understand the internals of the kubernetes but with a Q builder it gives you like out of the box a lot of utilities so if you if you want to make it dynamic like you get a request you create a containers on the fly then you shut them down right you you really need something like kubernetes otherwise you can just use docker compose which is what I initially did before passing you to Vitaly right so once you have your processes you can you know just use docker compose if you just want multiple servers and now with the latest versions darker actually supports native GPU computing right so if at some point we wanna you know multiply matrices right that's an option too so I'm curious because when I see communities and Dockers actually like a powerful tool but what I was wondering you know about the Google paper last year on the supremacy and then IBM came up and said well no we should be able to do this in two days on the Oakridge supercomputer of course the Oakridge didn't bother to give them two days of time to run it but i was wondering if one is using something like kubernetes and trying to distribute networks around the world would it be possible to try to validate or verify the IBM idea of solving the same problem it may not be two days it may be like a week but basically the same problem that Google did back then through a quantum computer and setting off quantum computer to validate whether IBM's statement is correct or what do you think about this is a simulator this it's not real computer it's a simulation of the IBM IBM used HP some kind of HPC to simulate the 53 qubits Mountain processes was the Sycamore quantum processor that they did the execution for two minutes they hold it and they generated two random strings and they claimed that basically this kind of type of computations practically impossible nowadays that they're right but I'd be happy I think IBM Emily I think IBM duplicated a computation classically with a simulator so they said even the quantum computation we can simulate in two days and a half right that's proposed but they didn't do the calculation yes so the thing is that it's still kind of hanging out there is it possible or not and I guess Oakridge wouldn't give that computer for two days to be running just for bet but the alternatives are using a distributed network maybe not two days maybe a week but still many if you can validate this idea that you can distribute it in a way that you can do it in a feasible time that's the question where it's doable or not sorry for so first of all a couple of things I didn't think IBM actually implemented their solution I think it was a theoretical stipulation you know that radical hypothesis that based on certain calculations they needed they could they they hypothesized that the same computation the same problems found by Google could be solved elastically even more efficiently than even faster than the quantum computer but I don't think there solution was based on simulating a quantum computer so in other words for example there are certain problems that can be found with a problem computer and with a classical computer but the classical algorithm is different from the quantum algorithm so if anyone with a type II maybe bacon but what I read was the point was even a quantum computation as is I can simulate it in two days and a half right so it cannot be or at least one of the points may be both points were made well the estimate that Google came up was a few thousand years but they were relying that this would be done actually in the memory of the computer and IBM said well you don't need to do everything in the memory you can distribute it on distributed machine and use hard disks and other things so you don't need all this memory to run it yeah but if you if you want to add to any qubits that would that's what we calculated with just the power you take the log base two right then you need a million you take the power if you need a million containers right so it's a matter of can handle can you handle that can you even the price is the problem right SETI at home programs men they have variety of fathers where people say okay I know my ideal CPU and GPUs to be used if somebody wants and that's a question of actually somebody setting up reasonable network can cook magnetic they pushed on different places and to perform the activities because it's kind of a enormous work still but yeah so when we talk to other people into simulators they assume you have to multiply matrices and use GPUs and swap numbers and so on and it was one of the points you actually don't write you only need to to use two-by-two matrices that's why they are that formula of a single qubit gate is a 2 by 2 matrix and that's all we need and it's a perfectly balanced like we said you have two servers talking to each other for a while one of them can do half of the computations because it gets the numbers from the other and calculates both numbers so they can send them back you know and said here are your numbers and the other server can do the other half like so it's a really perfectly balanced computation you couldn't hope for a better like Alexis said you know even for teaching a kubernetes I think it's one of the best domains you know actually if I might add something it was kind of a driver for Morgan Chase to open-source it right computer science a very good at benchmarking so if they have something I think this one has a great point because if there is something to distribute right then you can do a lot of fun things for instance you can ask GCP and iws to see who can simulate the biggest quantum computation and then you will get your machines for free most likely you can get credits you can that actually becomes a very interesting benchmark right it's a card distributed problem which essentially can spend any time on on running and so you know that would be super interesting to try and beat one you know I sure against GCP and edibles who can perform as a whole right better yeah you you can make it a service really if you want to but Google has its own simulator I bet right so I don't know how they do it but I'm sure they do it better with for themselves of course right but the thing we've seen of community like if command it boosts its minds or something it usually does a bad job so yeah yeah well what I'm hoping is that people who understand combinatorics for examples they'll see they can do quantum computing because they work with binary strings right you don't need or just computer scientists you need you know simple concepts and then once more people jump in you have you know a bigger community and from computational and black city perspective also what we might want to keep in mind is that adding one cubed cubed just to the 53 or whatever mod was used by this bubble supremacy experiment right we add one cubit that doubles our space that means you would it be for classical simulator or let it be a network of computer you would have to double your classical computing power now the number of workstation on this planet is finite at some point the complexity of exponential versus polynomial runtime you can't catch up with anymore this is why soon when when we do have machines with high enough fidelity that we can use a few more qubits to do some computations will soon run into a situation where it's really really getting really really hard to build simulators to to be to to to do is do the verification now add another 10 cubits and it's out of the possibility is there anything we can do it's just now that exponentially so complexity versus polynomial runtime algorithms unless we find algorithms that run in polynomial time that we don't know about yet so there is a lot of research to be done on that field I'm sure right so somewhere around 30 cubits that it took qubits maybe we we know we can do and it's useful right going higher you may have reasons to do it right but it may not be you you know a real reason to spend the money right like you said you double their computer with every qubit right so to add 10 cubits you have you need a thousand servers right so let's say we can manage 20 think we tried depends on the computation but we tried 24 26 on a single machine and then you to add 10 more you need the 1,024 containers right and then 2000 4000 right and so on so yeah at some point it becomes prohibitive it explodes but the good thing is that we can do it small spaces validations and once we are confident in the GUI tones then we can just trust the results in the impossible spaces where we cannot do classical calculations right thanks guys I think we have the unconference part and I think we have at least one talk by whistlin actually submitted I want to see if anybody else here wants to present or they want to ask kind of for a small explanation of some topics so we can have two kind of thing it's right you can have folks giving lightning talks for a few minutes maybe announcing projects inviting collaborators and another one kind of if folks want to learn something kind of 101 overview of a field they can ask if anybody here can teach us that so I'm hoping the flow for this if I think we have one talk by Veselin right which will go first and then I want to understand if anybody else wants to speak ins this is very informal no preparation necessary is just as we were talking about you know kubernetes scalability in the same fashion we can discuss all the topics so just checking if anybody once just say so just yeah okay so I don't see anybody yet volunteering so Veselin it's all yours then and if people want to speak after Veselin let us know okay so freely share my screen and then can I do that yeah okay yeah I can do that the slides to online if somebody wants to look at it and by the way for those of us that are in California I think we just had a earthquake here yeah with my connection that's a good sign yeah so I'm starting my screen share it says loading do you see my screen yeah I see here okay so thank you everyone and thank you for the opportunity to actually and present this conceptual again quantum computing for all I like the type of meet obscure running here Alex and I think this is well aligned with what I'm trying to do in the long run so this is idea where we try to build community around Merced because UC Merced is the newest UC campus here in Marzette in California but there is also the Stanislaus State and Fresno State which is the California State University system educational in the area and these are like a half an hour from each other from Merced with additional community colleges like five community colleges so I wanted to just share with you the local ecosystem and show that there's a lot of opportunity here if people ain't listened to engage so a little bit about we're set you can see it it's correctly in the center of California if you zoom in the map it's not from the Kremlin to mean it's close to San Francisco Monterey almost everything in Northern California you can reach it within two to three hours drive so as I mentioned they are different software developers I think we lost them the audio no the earthquake I think he may have been dropped [Music] can you hear me yeah we lost your screen for a moment yeah I think notice today a few times the system was logging again and again give my screen now I don't see your screen but you're you know you can just speak and while it is connecting because we can at least hear you okay definitely there are problems with the network on my site okay we can see your and him yep I can hear you too yeah okay and it's as it's loading the presentation can you see no I cannot no it's about the phoniest Valley some of the software developer workplaces around okay can you hear and see the picture we can only hear you I cannot see let me see I'll make your co-host maybe that will help you present yeah those Californian developments all the bandwidth you're saying they occupy all the bandwidth right so listen I propose that you know instead of sherry just tell us and hopefully okay okay let's start again and see if it fails again okay you can just speak with all the screen sharing okay I'm starting the screen sharing meanwhile looked up Merced it is indeed actually not far you know it's about an hour and a half from yourself so that's Central indeed Center located yeah when we have our airports around here too so do you see now the screen yeah okay so I wanted to point out here that there's also software developers water here many are moving to Modesto because it's much more cheaper and attractive to live and help your team in Modesto although your clients are in San Francisco in Fresno today is actually very big company bitwise that is very active they're building their own software developers and I like the approach to teaching computer programming because it's small like that mentorship and that's what I realize that it's Auto for some of us quantum mechanics talks years and it's a water passion but now with IBM Google in different quantum companies it could be something that people can program and do on their own and they don't have to spend years and years of doing it the standard way so it is to build some connections between industry and academia setting up kind of duo talks where people from industry would give a talk and then people from academia give a talk and that's why Merced is very nicely located we like for campuses University of California nearby then have another five down south this means that one can organize workshops in the Northern California in the different cups and these workshops could be kind of combination of this business academia interactions so that was our idea we were trying to organize such meetings and now I don't know probably one of the hopefully positive side effects pandemic is that people realize oh I can actually do this online so I can finish with a kind of list of people that are involved around here Tom Carter he is the chairman of the computer science department at Stanislaus we have J Sharpie Liliana the viewer she is engineer Jin Chen and few other faculties from UC Merced single that will do single tone he's from Fresno State and we have few others mathematicians and physics people there also in Fresno they have a business core that is more active so the idea is to involve students from the universities to assist in relation and the talks that would be given by industry and academia people and then business students to approach the logistics with the purpose of building startups that would be eventually doing a hackathons and growing into some business ideas related to quantum computing and I want to finish with UC Merced venture lab this is kind of an ecosystem that is really supporting all these efforts and if you wanna make a further feel free to connect with me on LinkedIn or send me email and let me know if you have any questions or comments I'm open for growing this kind of idea and turning it into a reality thank you for your attention Thank You Veselin that's a great review of local ecosystem you know initially we wanted to call this quantum West we open to the world due to kind of global nature of online but we're really interested in building up the west coast community so that's that's helpful I hope once we resume physical meetings the plan is to meet you know in Berkeley and Merced and Stanford I think it would be really you know Livermore it would be great to connect local local ecosystem so thank you so much yeah thank you anybody else would like to do a lightning talk not yet great well thank you so much it was a great second meeting for condo conversations we plan to keep going on a roughly monthly basis we'll put the video shortly on YouTube and the link will be also quantum SV and if you'd like to present it one of the next meetings please email me and hope to see you guys all again thank you so much thank you thanks guys thank you