sftext.org: Nikita Ivanov talks with Alexy Khrabrov
Recording: sftext.org: Nikita Ivanov talks with Alexy Khrabrov
hello everybody I'm Alexa crabber of the organizer of SF text and here our location at nitra where we have the first meetup of 2016 and we start with talk about that a lingua which lets you query data in each language and here also have Nikki Devon off and he's gonna introduce himself and tell us about how he came to build data lingual alright so my name is a kid I've enough just a few words about myself I'm I'm a founder of company called radiant systems founded company about 10 years ago was a CEO for 67 years in my mouth so our investor and a key advisor to date a lingua so very quick we did a lingua did a lingua is a what we call the serial like interface for your business data essentially going to use a free-form natural language to ask any questions about your data and get comprehensive deeper tones back that's the that's not sure what it is and so green is a big data company right it's a memory fabric right so can you tell us you know what kind of technology using great gain and what kind of technology use data link even kind of kind of how did this ID of the de lingua is connected with what you have done with creaking well greg n is the in-memory high performance fabric data fabric store in logic sets in ram and that's actually technology that has been used in de lingua fool and a high-performance don't be processing but there's some how different the the idea for the lingua came from kind of realization that you know you know last maybe five to seven years we have an explosion of SAS applications where you know every company from small startups to logic corporations tend to collect a lot of data most importantly keep that data in ever growing number of SAS applications you know almost any company when your size would have at least one CRM for like sales force in variety of marketing automation systems like cell google analytics and we're getting a hot spot in management like GRA and select for communication and something like workday for charlie is going on and on and that's literally for a small start-up you know larger corporations companies would have dozens of the systems and the data is basically completely compare meant alized in silent and that's something that we've kind of fought in our late 80s early nineties and we kind of back to the same story where we collect a lot of data and its really isolated and the company typical business would have one or two folks that actually know how to use each particular source mm-hmm but there's almost nobody who could basically have the same knowledge of how to query google analytics and salesforce and hubspot mhm and therefore data has not been asked in a proper way it's very costly to ask economy no consistent report across multiple data sources you need to involve engineers and typically it stretches the days and weeks and it's very awkward hmm and if I think about how do you solve this problem what's common between Salesforce and more care there's nothing nothing in common between you do two systems is complete different data sources complete different data structures they can text your company for the interface about the only thing that is unique that is actually shared is that it can actually ask the question about two of them using natural language mm-hmm I mean you don't have to be a specialist in Salesforce to asking what is the average sale from east coast of those months all right you don't have to be marketing professional to ask you how many web page has been you know viewed in our website for the last week mm-hmm see those questions can be asked almost by anybody yeah you know if you would go and start learning google analytics and sales but it will take you weeks and weeks and weeks of learning to run those reports yes that's the fundamental idea fundamental kinda know the backbone idea of data lingua why don't we basically shortcut whole thing and if you ask the question in normal language speak or type mm-hmm we're going to give you back a transfer mm-hmm and today I totally will launch the private better with one data source for google it exceed a later on and you can literally you know use normal language speak or type and the system will respond back to you this is this is awesome and I you know had their kind of peak sneak preview of this right and I think it's really it's really amazing and I agree from cadman from big data side and seen a lot of companies in Big Data special think specialist mobile devices natural languages the best way click on write as variants you know we kind of in NLP community will Kennicott excited thinking that like natural language is again sir for everything right but when it's really the answer for almost everything ticket to talk about exactly mobile you might look at the SmartWatch look at this phone's right here do they have a keyboard no and five ten years from now it's kind of long horizon but you know we'll look and they're looking our kids today then I'm going to know what the keyboard is look at the navigations in our cars five years ago you were painfully typing it in for some awkward keys on the dashboard today most of the navigations you talk to them mm-hmm we're going to have exactly the same way this this transition from the consumer anal p to the enterprising loopy will be very much overnight because technology is there we just basically have this like a little bit between the consumer enterprise and especially with a prefiguration of multiple data sources and be all them being completely incompatible there is absolutely no way look at Syria today I mean if you have the Syria or Google now or Cortana from Microsoft nobody in the right mind would go in a complex search for lettuce sports scores or whether is movie recommendation you basically asked yes it's dramatically quicker yes at least for this specific data sources the system is already very much better than you can basically search manually yes Anna just in the last few three to four years we're going to have exactly same impact on on the business data you are will you you will be able to ask questions simple and it's actually fairly complex I can data link to it and it just answers back this is interesting because I just seen somebody talking on facebook that you know 10 years ago a newbie would come to the web browser and would type a question into the URL a box you know what is the capital of thrill and thought they're idiots and now actually this is works like they they try to look the try to do right the system's adopted right on the system is basically sir and what they try to do right so so I think this is this is great so how does should a little bit and kind of talk about how you came to do this and also you know had a kind of bluish look at the data codebases some point in the past and I was you know as a scholar guy I must say this is like one of the cleanest best maintained code bases and you actually wrote a lot of the cold yourself and we also work both in Scala and in JavaScript and and I'm you know and you're basically you know help and drive to companies I just want to understand how this is possible why did he use scholar can you talk a little bit about the development right so you you you have basically kind of active from the development of day to anyone how does it work for you and and kind of what's most interesting for you to build this system from scratch well you know the idea their lingo came to me when I was acting CEO of radiant systems so obviously cannot run two companies in parallel so what I did you know I founded a small team in st. Petersburg Russia that you know basically spent last next two years prototyping this in different ways we started with the computational linguistics it didn't work we moved in the deep learning and neural networks in double in this direction didn't work in eventually way you know settled on what I'm going to talk today like a grammar semantic grammar mhm and that space is fairly fresh new I mean there's quite a few research going on like an apple by the way Microsoft but essentially up until very recently it was just you know having central research project with you know two or three people sometimes we hand it up to five or six people in a team mm-hmm and you know haven't run great game for many years like I kind of learned how to do it you know in a very economical new bootstrap fashion mm-hmm we use a lot open source software you know the typical infrastructure today for a sauce company is very inexpensive to develop and run and lunch mm-hmm and it's still a no darling who is still very very it's still a pre-funding stage which I kind of as an advisor is a kind of a founding member of a team I always recommend that you know your delay the fundraising Tildy yeah absolutely necessary have to fundraise and raise the to scale the company at the scale to actually launch properly but I think that only way today is in the stage were there is a working out public version in neutral version there is what we call you know minimally Viable Product out mm-hmm it's rapidly gaining traction its life everybody can sign up for free yesterday so I think the next six months for the team would be to really listen for the feedback from early users mm-hmm to really you know quickly iterate over the product market feed mm-hmm and get to the point where they will be ready to fundraise yeah this is I mean I'm pretty sure that in the most folks see the presentation don't see the full potential see what a little bit o develop it on your kind of right and your ability to to be in the code I mean like a lot of executives are not able to do this right i mean and even if they did this in the past they kind of move away from this so you so you're actually able to do this and it did in a very rigorous way so I'm just like I I in our meetups we actually like this angle right like we can where we get a lot of technical founders and and I think especially for the platform company's cutting edge interfacing of a lot of systems you need to be technical enough so so it looks like you know like Lucas Park most co-founders are actual open source developers right right and authors of the original system and and so and and then the system becomes extremely popular among other developers right and so I think that like kind of a thing with a patch ignite I think you're kind of right for the same All Right see the same trajectory so so I'm really curious if you can shed light on this because it's not very typical in certain not very common we get in our technical founder and also I think you're spending these two worlds right you're spending the back-end and front-end and it's usually two disparate worlds extremely hard to find and in but that link was beautifully done right that's kind of I think it's it's really you were able to build the UI which kind of you see it's very professional in the your bicycle thing like it was built by a huge team of people right like there is you know a lot of details in and kind of bootstrapping this and with the startup so I wonder you know how you were able to do this and be in a technical founder what like how it helps you know in the code base to guide the technical project like this to completion and what would you what we would be your advice to you know are the founders it kind of if they want to stay technical a bit more technical right like what is the importance especially like having a technical co-founder being in the cold being close to call to to manage and and build the system like this yeah well that's a kind of loaded question i agree yeah but i can tell you my kind of eunice having you know run the company if you know raised millions of dollars in venture capital there is there is no non-technical founders mm-hmm salespeople never start anything mm-hmm marketing people never start a company they may think they started but basically they have no idea which is always wrong and they are technical people that basically proved them wrong the only people with almost absolute certain that start the companies today especially in our area here are technical people mm-hmm now not every technical person could be good manager could be good c or could be good salespeople and it's perfectly fine but you know being technical is almost prerequisite if you're not technical if you're not technologist there is no way can start anything mm-hmm very very rarely even if you hear that the company was started by salespeople you just don't hear that a story you just don't hear the other people who were involved in very much the same thing it because you know salespeople and nothing against salespeople by the way there are extremely important in it so the point of company I have to sell the stuff exactly but initially you know when you build a product build a service it can be done only by technologist so that's going to answer is one of your questions how do you keep yourself being you know in executive in the company in raising money and the management sales and marketing team and still quoting you know what I like what I'm doing so you know I always always run two or three projects in peril obviously there's a one project that technical prize most of the time like green game when you run the grogginess see or obviously better Capasso most of your time but you know if you're good with the timing if you have a certain level of productivity personally you can definitely have a two or three projects running in parallel mm-hmm and doesn't mean it has to be involved in every one of them Oettinger percent but you can establish a certain organization established certain you know engineering discipline mm-hmm it's very important to you have an eye for you I and studied constantly the new I concepts and you experience concepts and how the new way I is built and I do that religiously that helps me basically drive-in comment on the UI and ensure that you know you why is done the to certain level which will like but again it's you know there's nothing you know earth-shattering here it's you know I agree with a lot of developers kind of shy from you why because it's a very different you know mindset mm-hmm you know it probably helps me I grew up in a very different you know unusual family my father was one of the largest collector of art in Russia mm-hmm so I grew up across the paintings i grew up in cars and shapes and formed from very early age so it's kind of near dear to me both life photography yes which is also very kind of geometrical colorful we know semantical world if you will visual visual in many ways so these things do help but it also helped probably more than this is to involve people who understand the same concept because you're not going to be doing everything and it's fine it's it's easy to do when you company five or six people everybody is writing code pretty simple as this when you grow up to 15 20 people not everybody at the writing coming yes 30 40 forget minority oh yeah yeah as a CEO you basically char person company so so guess what you know yes hiring the people initially or involving the people who share the same outlook and what good engineering is bought but on the backhand in the front end is critically important great that I mean this is some great insights and you know on this kind of you are a theme there are so many details typically you know they're all of p.m. emerged and so the role team was a full on top was a mystery to me until I came to an you know I kind of have a hobby kind of asking different organizations so I've been at Microsoft I was in the cafeteria with developers and asked them what do they think PMS do and can they overall mystified and kind of they said like they're running around with Excel spreadsheets if they report their other senior PMS and kind of em so right but the startups is crucial so I've met some really great p.m. right and it's it's another difficult level to have PMS the data space because they need to understand various aspects so I wonder do you have p.m. in your team I you your own p.m. like what would in order to shoot this you I most startups believe you need to have a very strict p.m. / wheel drive every feature on some kind of compound system so how did you do your p.m. yeah I think actually the need for a product management is clear I mean perfect for anybody right now it is the complete to hire it mm-hmm give example a great game the company is almost ten years old we don't have a BM hey we tried multiple times it's a very because the PM is the kind of job that you know runs triage between sales marketing engineering mm-hmm it has to be very independent has to be technically minded and has to have an aesthetical marketing sales it's a very hard position to hire naturally when a young company when it's a youngest like sea level 0 pre-funding company you don't need to be out you are up in founders or a product manager I guess if you don't have an idea what you're building you don't have to be in it's something that's wrong but p.m
comes into place when it's very you know if you're what i call if you've seen the movie in developing a company / over there / the time you you will sense because when you're going to start seeing a lot of friction between sales marketing in your new rank mm-hmm that is not resolved by quick emails mm-hmm and you constantly run in the meetings and you know and and intentions fly that's where you need a product management mm-hmm because typically come in a start-up is due around for quite some time with founders being the PM's mm-hmm because that's most of the ideas of a product or service came from but rapidly and typically it's anywhere when the company grows over like 50 people something like this when you have you know five or six people in sales organization you have three or four or five people marketing in 2025 and engineering that's typically where it starts start to see basic a lot of friction and the only way to isolate that friction is to move this function to a person mm-hmm who will work very closely with engineering understand features but drive basically drive the the essentially the game between these three functions of a company sells not getting an engineering it's a very I think it's wrong to think that you know product PMS or will somehow develop the features on their own mm-hmm it's a kind of secret you know who it's typically not it's typically uh you know sales people will come from some ideas from clients the marketing will come some ideas engineering will come with the most of the ideas the Java OPM is to really kind of corral this information the proper way mm-hmm and be this kind of mediator between you know three you know equally important you know parts of the company mm-hmm so having the function of PMS are pretty simple it is extremely hard to hire them do you think they have a certain master like I mean like what all these three things out there really should be equal or are they different in different companies like it do you know p.m. effectively convey you know others or ideas from management of some kind right down to the engineering or you know that that's the wrong way to look I mean obviously you look at the bigger companies it could be any kind of you know setup but in a in a small start-up typically it's a technical position mm-hmm first and foremost because typical engineering teams are much bigger mm-hmm in sales and marketing teams listen usually mm-hmm when you're talking about the small start-up so we have to communicate well with engineers absolutes because typically in the typically p.m. start taking away some of the mindshare from engineering mm-hmm because typically it will be marketing sales complain about engineering mm-hmm that something isn't done on time we're doing something else for it or not the giving us features or something is broken right so typically marketing sales complain and engineering engineering basket tells you will live us alone we're just ranking up the code here and the daybed daily basis right so this you know productivity cool person will comment in start manager in this process mm-hmm so that you know there is somebody in the between so the people don't conflict talk to each other directly but there is obviously there's a conversation directly communication as much as you like but there is a person whose responsibility to manage this relation between the three aspects of the company mm-hmm and it gets more complicated you know you get you know product you know you get you know partnerships and you know and other things going on you get professional services going on all of a sudden you get you know your marketing grows dramatically over time different factions so PM has become much more involved and you know you grow over hundred people as a CEO of a company you don't even know what in your meaning does you only interface with them as a product marketing people Oh product management people right so that's good because you know you you can only manage so many things yes and you have to constantly hire the best and isolate itself it can basically spent quality time management functions you can lunge but always saying the good the best job as hero can do is to hire the best people get out of way mm-hmm and I always say another thing that you know when you build a good team but you can go on vacation for eight months nothing really happens bad and now it's got you know sad in a way because you're not important anymore on a daily basis and it's four founders it's very hard but if you go for a couple of weeks away and nothing bad happens and just business continues to go as planned daddy know you have a you know a minimally good team in place you know this that's a great site but I mean in your case it may mean that you will you know go and heck Celtic and skull and angular right on your own so I'll probably kind of lino close with this question right because you know also members of escala meet up kind of overlaps with this text a lot typical choice kind of for good choice for for back end and but you also build your front end and angular and has a very interest inside that angular makes it easier to manage a project then say react right so which I find extremely interesting because it's not something typical using comparisons of reactor single so maybe you can kind of explain why you chose Scala and angular for the stack after that I think about and how do you see are you sticking with this going forward how is this working out how do you think is going to scale you know and we'll check a prediction of scale next year or next time so if so Scala is very simple i laughs Carla I you know I haven't touched job in like six years I don't know actually remember how skur jaw looks mm-hmm very ugly probably even Joey probably doesn't doesn't make any better so scallop was natural choice it's a great language you know languages languages you know we can use any languages but you know it's very productive to thinking scholar Lincoln's you know in a stroller type environment which is very functional in degree but not I know as opinion it like Haskell for example and still right in the old pea coat when you need it mm-hmm and it can be functional when he needed when you like it so I like scale I like the practicality of Scottish you know I started in many functional languages nothing stick to me but I like Scott because it's a very practical it's done surprisingly is done by you know guy from academia like but it's done very much an engineering profession mark moves between academia right so there's a lot of shortcuts and I like that it's a very practical language it's not you know if you look at Haskell it's much cleaner but it's a massively opinion it's hard to build anything beyond the toy projects on yes which is another problem that's impossible there's no point in libraries yeah the community right this is inherently limited exactly as far as angular when we I remember when data link we actually started in the by the way grid game project house it has a and go GS based UI components as well will often go GS from one for the middle reason you know if you build in some you know halo worlds right anything at work you can write it in a jquery directly mm-hmm react works perfectly fine media or works like you find you know angle GS work work define it is when your project grows quite a bit significantly a special angle or one expo you know a sudden structure of the projects with controllers with Hogan ABC model maybe see like model mm-hmm which in the beginning looks like a constrained in awkward too much you know too much boilerplate mm-hmm but the miniature actually grow into the you know five ten control lists and 20 services and whatnot in templates Oh a sudden start to appreciate that neither society it's a massive ecosystem it's a massive community it's probably number one in a front-end framework it works beautifully with boost Robin html5 compatible obviously everything else so it's it's a perfect way to build systems like that lingual gets going to show you see today it looks solid you know everything is built in an angle GS and just knowing by people who build it what actually surprised me surprised me a lot of couple years ago is that the development in javascript is very close to developing in Scala it's a very functional especially with libraries like lodash that gives you pretty much exactly the same you know you know it will routines if you have in Scala it's a very very functional approach jealous keeping itself is kind of a little bit dirty language but it's functional in the nature mm-hmm although its net doesn't happen you're stronger because Lee yes very loose in you know and there's no strongly typed anything there but there is a keyword let alone recently in the new center which confused me with a camel for a moment yeah I'm not I'm not going there but if fundamentally uh you I was surprised when I looked in the co2 is that if you have a strong functional background it's very easy to jump back and forth between Scotland JavaScript because very much you know the same thing supply but there's a lot of differences in you know and syntactical things and whatnot and in performance optimizations dramatically different but I think it's good i think it was a lot harder to jump between java in javascript them between let's say skull and jaw ski which is very was quite of people you to me interesting interesting so anybody who's writing the scala unless you're completely puking about in know you're white work it's a very easy you know kind of straight back and forth between you know solid javascript work and especially people like work with a calf for example because javascript is fully synchronous there's nothing synchronous in JavaScript it's a one-third of the environment you have to constantly work with futures and what's called the promises in JavaScript mm-hmm so all this background from scowl developers comes extremely handy nice so it almost looks like we are looking for you know acha HTTP Anaka streams are almost ready for the framework which will connect angular 2 you know acha streams on the back end and then everybody's very reactive there's a socket i/o there is it you know this this world the email look you think Scala you know ecosystem is full look at the NPM look at the JavaScript ecosystem there is like a tongue at times more libraries and everything is available yeah front end the world is swimming in choices so hopefully maybe you know we'll get some time we are looking for somebody who'll teach kind of and you know angular face color developers right maybe got somebody but you know I think this this is this is this is really exciting and encouraging so I hope you know we'll see more full stack people will be the focus of this year we'll have a hackathon later on and sense of scholar well you know build full stack applications so you know anybody who wants to learn hopefully you guys will join it'll come thank you very much nikita it was great we have a lot of more questions but we'll save them for the next next meetup talk or a conference and we're looking forward to your presentation tonight thank you all right