Transcript: SF Scala: Vlad Giverts Interview

hello everybody I'm alexic rubber off the organizer and founder Joseph Scala and here we are at Clara the first time it's our last meet up of the year and we kind of start with the new company this is we're very excited and here we have vodka verts with us who is the head of engineering and Clara and he is a loner friend and partner of subscribe he hosted us before Thank You Vlad and we're happy to be here yeah happy to have you guys here thanks thanks was fun thanks to the wetness so woman I think at least two different previous companies yeah we met first identified you invited us and we shared the common law for intellij idea which was the idea of choice was Carla and still remains the case right and then you where at at work day mm-hmm right doing Simon doing machine learning and building the group there and so then you are kind of inviting us to this new place so tell a little bit what is the trajectory which is this called connection you know what is Clara what's no its new yeah so I mean what is Clara is probably a bigger question so I'll go into that probably a little more separately ok but it just in a nutshell why does Claire exist and that is the vast majority Polinsky do not fully understand how our financial asst to works yes and they are not empowered to own their financial lives and make the right kind of financial decisions for their own benefit mm-hmm so what we're trying to do is to empower them with kind of both education information con transparency in our part mm-hmm to take control their financial lives and we're starting with mortgages mm-hmm scala connections interesting it's actually some of the early employees kind of want to basically out one of the first of engineers at Claire it was an ex Twitter okay so kind of had that Scala lineage he brought it here we run on a Twitter stack finagle all of that all right serving us really well all right for a financial company particular like having very very strong types making sure that we're not making you know subtle errors that can result in like tremendous cost to us or our customers mm-hmm and Scala is very helpful with that cool so I didn't know about the finagle foundation because you know we just ran Scala by the bay ed yeah right where Finnegan was in the air right and it was the second time we actually did finagle at scala by the bay right Flossie original con and then we kind of do the whole thing together and so one of the reasons we want to bring Finnegan out of Twitter more right and want to help them basically to kind of bring the goodness of it to the world and we're getting a lot of value Ataman we have a micro service architecture mm-hmm and basic finagle is kind of infrastructure and the glue for all that all right no this is really cool ah so neatly given a lot of ideas right how to kind of connect more people to this so but tell me a little bit about kind of your trajectory in engineering right so so I mean you've been identified I know identified since they were on Ruby people they switch to scala right yeah so I kind of follow them and and obviously workdays a one of the biggest users of scholars they send a lot of people to to scale by the bay right so and you do like the photos part so can tell me a little bit you know how kind of your career kind of a line of scala what you found useful what you found hard what do you want to not repeat a car what they want to do better at Laura sure so my first experience with Scala was probably in 2009 when I was at a company called tag tag the time was the third largest social network in the US after first was my space then Facebook then tagged and one of the other engineers it was into functional programming he trot who's trying ups golly he rewrote one of our services in Scala go on I was not Johan he was the CTO that's right this was one of our engineers eat I and Johan the CTO look to the service at hate this is very elegant you know what was several hundred lines of code he rewrote in like 30 or 40 s it seemed to perform equally well I was much easier understand and he said no we can't ship this mm-hmm like nobody here knows Scala there's no good tooling support for scholars no idea supports like I'm sorry this isn't gonna work out eat I was disappointed I was slightly terrified of Scala could have it you know kind of a simp similar reasons mostly because it seemed inaccessible to me so we didn't do anything with it then fast forward a couple years I'm at a startup called lion side we're facebook social gaming and I was had joined as a very senior engineer head of engineering VP madryn didn't work out I somehow became de facto person running engineering over there okay and I was an intelligent user and IntelliJ Scala plugin came out yes so I said all right let me try this out and this hadn't just come out I'd been kind of playing with it a little bit some new iteration which came out and we tried it okay we finally have some likes decent syntax highlighting and a little bit of auto completion actually showing you this red squiggly zhonya's know where you make the errors yes it's all right on the right okay look this is actually viable like I think people this gave people it felt that it was enough of a guardrail for someone who's coding a job at a try Scala out and even if they don't know the syntax the IDE would kind of guide you yes so we tried it out and we started migrating some of our Java code into Scala it was nice okay and it worked okay and it was slow and it was kind of cumbersome but like it did the job and we were happy with it fast forward another year and a half I'm now CTO at a start-up Cole identified yes we're hitting all kinds of scalability challenges and we got there's got to be something to make it run and we were using a bunch of ruby on rails which is great but like you can't write like persistent services that are running and processing things in memory like no you got like you rut essentially it's almost like a PHP model rather start a process you do stuff and you're done so we we tried Scala for the I knew Scala nobody else in the company actually did but I kind of encouraged some engineers to pick it up they ran with it we're really successful nice and they did a lot of graph processing right so this is kind of hard to do in Ruby right on the I mean tutor switched away from ruben foo yeah same reasons yeah so then we started processing our big data in scalding mm-hmm Hadoop yes and scalding was a really nice way of doing that I'm not gonna her office or something yeah yeah so that was that was really helpful then we get acquired mm-hmm by workday yeah and we kind of became the data products team at work day because data analytics machine learning and we went from like a half Ruby half Scala shop at that point to one hundred percent or at ninety percent Scala whereas all big data all the time all the services were writing in Scala we'd moved up over to spark mm-hmm which is all you know Scala base yeah and I kind of never look back and then incidentally when I was looking for transition and I chose Clara as my the next step for my career they were already on skull all right so I got all right all right right yeah no this is great this is great so Simon still works for Simon team which was the former identify teams to work this they're very happy and growing yeah we you know we've met a lot of folks you know what one of my good friends works there so that's yeah that's that's very good so kind of Ewing include amount of Scotland startup world and kind of the benefits a comeback I would kind of give you a kind of an interesting data point and I wonder what you think about it so I found that in the last year yeah several fin tech startups wish wish I kind of called loosely protected by anything so basically we should have something do finance start using scholars so there are several companies we should be using Bitcoin for whatever reason uh you know Bitcoin blockchain actually has a lot of open source projects using scholars own so that that actually is interesting it then there is not a communicant alley and they recently joined us the funnel by shasta which finds you know a lot of scholar companies right including website flight band and so they want to build a new bank they basically want to effectively refine your credit cards right take a better line of credit to repair credit cards and let you manage that mm-hmm and when i talk to them what was the reasons they chose scholar they mentioned several of the measurement correctness right I got the financial situation but they also mention that it makes it potentially easier ah to enable acquisition because financial com is no job a lot of banks are using Java right so basically the the already of GBM and so the reason goes that scholar runs on gdm at the four it's easy to interrupt with larger companies and so I wonder if you know first of all kind of mr. present to me right because you think finance of their conservative industry but here's actually an argument for legacy compatibility 3g am right in in kind of a technology which would enable potential integrations with a bigger companies i wonder if you guys thought about that or if not you know like what is what do you think about this line of reasoning yeah so in our case we really really do not want to be acquired mm-hmm so we're it's basically transform consumer finance or go home mm-hmm but the reasoning makes a lot of sense so I know when workday was looking to acquire identified they were on the JVM stacked running mostly Java mm-hmm we were running a lot of Scala yeah so for them there's all this naturally slots in we know how to run Java we've got a lot of JVM expertise so if they needed to transfer people to our teams that's something they could do so definitely played into their calculus for that acquisition so I'd make sense that other FinTech companies if acquisition is a potential outcome they're looking for mm-hmm and that's something you should consider yeah and I mean I don't mean that note Ali wants to acquire that's the reason I write like I mean obviously everybody wants to be yeah very big and I think another question is I'm just curious can feel you know a business model right that like you know to play here you lot of capital yes right so I think if you want to be a bank or if you're going to facilitate mortgages connect a little bit more how do you add value to the horrible mortgage business I just I bought the house last year's like basically I mean to me by the house is just my cousin PDF right basically it's PDF and electronic signing and and ridiculous credit approval through slow-moving banks right which is so like it's broken many many levels so I wonder you know how would you want to prove it with all the technology yeah so it's interesting mortgage banking is really an information processing problem mm-hmm because what what what does it mean to get a mortgage it's kind of like filing your taxes that's kind of filling out the mortgage application yep and then getting audited at the same time correct so and that's not very fun yes what we're trying to do is streamline and automate as much of that as possible so instead of you getting on the phone and dealing with your broker or loan manager whatever they were ever they were and concentrating so what's going on what's what's happening what do I need to do right maybe that maybe sent me up there calling you and telling you what you do instead imagine just an online experience mm-hmm where you go step-by-step you come well collect whatever miss you need you put it all in there mm-hmm anything most everything is automatically verified via API integrations mm-hmm and there's all kinds of complexity that today is done by you know got processors and underwriters and compliance people locked desk you know capital markets and closers funders I mean there's probably like five other functions I'm not mentioning which is incredibly human labor intensive yes all of that being streamlined orchestrated by either workflow systems or whenever possible automated and as a consumer you can see what's happening step-by-step so it's totally transparent to you mm-hmm you know when you're more just gonna get close you know what the bottlenecks are and you always know what's on you mm-hmm so the other day what was this horribly painful process where you have no idea what's going on or even if you're going to get the mortgage this becomes like open and transparent yes and so heavily automated it probably takes a small fraction of the time can you do you plan to give her guarantees over closable time window will take eventually yeah we're not there yet because that's a thing that's very important right in the area right they demand very short closing windows yeah and then the banks could not deliver it the rest of the traditional banks a very slow so i'm gonna use traditional banks and basically again i used underlying banks and for financing or your own so sort of hmm we're not inventing a new model for mortgages we're not creating new sources of capital and some players in space are mm-hmm we the way it works is we partner with warehouse banks kind of like wholesalers for money yeah so when we close your loan we have a credit line with them it would literally just wire transfer the money from our from the warehouse bank to your escrow account and bloom you have your loan you have your home and you can move on mmm-hmm we then sell the loan to seller aggregators mm-hmm these are the these are companies that purchase these securities these loans yes from other mortgage banks right and then they sticking bundle them into what this coach or Julie she the mortgages what's that that so like this Sora composition shade mortgages right now they are something quickly to exactly it was aggregators exactly and eventually will become the aggregate ourselves as we get big enough sui will dissing me disintermediate them we've already disintermediated brokers we're doing that part like talking to the customer directly ourselves hmm and then we'll eventually sell directly to the capital markets mm-hmm whether that's Fannie Mae or Freddie Mac or other investors but we're going to interface directly with them so the goal is to create a short of a distance as possible between you know a home buyer consumer and the sources of capital nice so you know I person found like red so the problems with banks with mikey is right like no let's say an experience is our own social security number mm-hmm for my wife the completely stops bank of america yeah the basically said go back to experience fix it mm-hmm right and then proceed yeah and somebody told me Bank of America will never close in time and I said no they give a very good right I'm going on dreamin customer yeah they know they won't do this right they almost cry they asked me to consider the fact that they will not be able to do this I feel extremely short closing time so I just for fun just to entertain the humor this person I initiated the process with two mm-hmm because and I felt bad about this because one of them is not gonna get the deal right they're gonna work for me and not i'm gonna say no you know i'm with the other one because you know maybe if there is a different and bank america failed because experience have there in a bureaucratic mistake that's completely substance on that tracks they say go back to experian you cannot even call experience yeah right there's a whole bunch of you'll experience a mic reports is wrong give me credit I'm mccray score is wrong because of some states right so so basically there are very fortunately the second worker was able to do this right so so it was very grateful to the person who advised me right so the traditional bank apparently post-crisis right there basically super conservative so what are right but you know the infrastructure is he right there is transgenders experience right Equifax and so apparently for bike market secrets this regulations mm-hmm you know numbers of the match two out of two much third doesn't match go fix it right the basically right so so I were able to to fix this kind of errors right we should throw a monkey wrench into the whole process are you are able to deal with this all right because if you're one of the banking bank itself right back we'll do the same thing even if you do it online so so I were able to apply the intelligence right like an intelligent person looking at this probably will say this should not stop right my application mm-hmm so how can I deal with this kind of stuff ish current people perceive as kind of bureaucratic nonsense that's a great question so one of the things that we're trying to do with our workflow systems our automations is actually leverage people where people or what people are best and that's applying judgment mm-hmm so whether it's this case or some other like non happy path case mmm that maybe the automated systems or the workflow systems can't handle mm-hmm we head will have an exception flow mm-hmm where computers handle the easy stuff and if something's out of the ordinary it goes to people mm-hmm whose job isn't to make a routine decision but actually to exercise judgment mm-hmm which in this case would be a perfect example of okay like obviously this is okay there's just like one out of three is off let's just ignore them and move on yeah and it's not the primary right so it's kind of so-so but that will be possible because you will have your own judgment about this warehouse bank credit yeah well I think there are certain restrictions that's like what Fannie and Freddie Mac are willing to deal with mm-hmm but that's not that's not one of them so if it's purely internal restriction mm-hmm then we're actually setting people up to exercise judgment rather than setting people up to follow a strict process hmm I think it's kind of just in terms of how or how we're organizing ourselves that's quite a big this so it looks like you should be a mortgage company it sounds like you have a lot of machine learning to do right because some of these things I can be scary yeah so whatever plans from ml and machine learning and artificial intelligence you know can you become a guy company mortgage space interesting so we are going to apply a I it's going to be around a few different things it's going to be around customer lifetime value mm-hmm try to unfinished I what kinds of people should we have put more emphasis on it's going to be risk analytics hmm trying to gauge what is the credit risk that any given person poses that's actually going to be very valuable to us if we can demonstrate a superior risk model based on AI we can actually market our mortgages to investors for a bigger premium you literally have higher revenues if we're really good at AI mm-hmm so that that's probably the most immediate application interesting and how far are you from this that we're still a ways we're probably going to start doing our first predictive analytics more on the user acquisition side like more marketing side of things and maybe q2 q3 of next year okay and not too far away not too far away lifetime and start up yours yes yes yes yeah and then more on the risk modeling probably late next year early the following year okay and do have any idea which sticks which software sex again use for Marceline not yet okay not yet all right captain water down adopts we're gonna write a post one of them we will cover by area in addition to the as a scholar and the some spark right now so yeah so I'm a lib and sparking I'll you know it's all kind of good options at this point yeah cool so super exciting looking forward to your talk about chloride or when meet up and thank you very much for foreign skier all right thank you Alexi what