text anoki interview 12 15 15
Recording: text anoki interview 12 15 15
how everybody I'm Alexa crab off the organizer of SF text met up and here we are on location at smart news it's a new company in the space in recently in San Francisco a doing really great work original from Japan and here with us we have now aqueora who is an engineer working on the machine learning team hello kids it's thanks for hosting us here yeah thanks for coming in so this is our first time as partners can you tell us a little bit about Marcus what does it do we're the largest mobile news aggregator in Japan and but we consider ourselves more of a machine learning company and so in the back end we basically have a search engine you can think of it like a search engine where we have a crawler that crawls the articles on index says it and whenever the user uses a bit basically runs a search on our index and so there is a lot of machine learning tasks involved in this we want to categorize articles we want to do particles we want to rank articles of so a lot of machine learning goes behind the covers of smartness so this is interesting i just learned that you know you guys a big in japan and you have millions of users so you're a recent entrant in bay area space we're very happy to connect but you know i'm curious how how does it work like what drive this big success why do you have so many users what keeps them coming back I see um so in Japan a lot of people commute using trains so they basically have a lot of flexible time flex up even if they can use with their smartphones so I think smart news in Japan at least it basically is a way for people to spend time on trains or during lunch or basically free time to instead of playing mobile games on smartphones it's its way to basically check the news ok and you know we know you know news can be very negative right like if you open a newspaper there is terrorist attacks there are shootings there are Wars right so I don't know you know people do check the news but what keeps them coming back to to your application how do you tell all right so like people do they get positive experience like can you find the news weather which are kind of positive you know bruin alive like what can you tell us about the character or the news you are feeding to each person yes so we try to optimize for discoverability so we need like normal on use applications i think they tend to overemphasize personalization and i think that makes sense for for example when you're buying stuff you're interested in purchasing things so you should the feed that you should get the highly tailored but when we're talking about news on certain events if it's really important I think our philosophy is that it should be delivered to everyone regardless of your interest so even if you're like an LPG living in San Francisco yes and you're interested in NLP and Scala and spark we still believe that your feet shouldn't be hundred percent those kind of stuff yes there should be stuff about presidential elections yeah cars terroristic uh-huh and so on okay interesting I will just have a wild idea and you can completely kind of say it's wrong right but I was thinking you know I'm always reading all this negative news if I have a power to create really personalized app I want to find news like completely like I won't leave my own world let's say I want to have news about people let's say who learn a new you know new language people who learn your craft you know people so can I compose a new stream which will be completely you know my own world it will be about positive things it does not have to be you know all about my interest but can i say i don't want to see this news about remote wars because you know i have no control over them they i cannot do anything about them like can you do something about that or do you believe that like no like it you this would be you know global news and some of them are negative we're gonna give you anyways can I can I turn off the negative news if I you know I'm sure they have this power so right now in smart news I don't think we can do that it's it's kind of leads to this filter bubble problem like i said earlier so thanks it really be aware of the tourists that that's going on right now and with the thing that you just explained it right now because the tax are so sad like it if we've just filtered them up from whatever feed that you're receiving i think it just um leads to too bad results and and and how about I hear about it once and then I say again like I got it like okay you know thanks all right yeah but that's a good point so I think in the end I don't think you want to focus too much on it's basically like you want similar to when you're eating you want a well-balanced diet right yeah so I think for news it should be really similar you don't want to focus too much on a few things you might want some sad things like maybe twenty-four ten percent of the time and that should probably be like consistent throughout each day true yeah sure that's a good point so I've seen you know you have a closure or log on the laptop you know and I also organized a scholar meet up in our like closures the sister functional programming language so I know you guys are on JVM but you played with spar so can you tell me a little bit about the technology stack and like we know your experimentation with spark what are your plans in that space yes so our entire stack is a basically where Java shop and some parts you spark so obviously we used Scala and our entire system is built on top of AWS and we have it built so that um basically we're on this this trend called microservices vision so a lot of our services we're exactly using micro services so engineers can focus on a particular sub component within our system and they don't need to spend time like on deploying stuff or debugging or or test well we do deploy and above it's a de pointes automated but the basically we try to make it so that engineers can spend a good chunk of their time focusing on core algorithms nice nice and so are your where is your engineering is it mostly in Japan or is it also here so right now we have two engineers located in San Francisco one including myself and we have plans to hire more in 2016 16 but right now i think there's about 25 to 30 engineers based in Tokyo mm-hmm and so is it you know easy to hire let's say people who do machine learning and data mining scale because this is some of the higher hardest positions here you know it's it's really what is important right now scalable machinery data mining so you know it's very hard to hire people how this is the situation in Japan like you know is it possible to hire these people obviously you have built a very good app so you found a good people right right right how does it work yes so in Japan I think because it's it's we have a pretty good user base it's well known and so it's relatively easy for us to hire good engineers but in in the US it's still relatively minor so we're still having some pretty difficulties so if anyone watching this video is interested please feel free to reach out smart news com and I must say again if you know for watching I think these guys I really did the right thing they invited us to host them it up here we have dozens of developers so this is the best way to connect with the right developers hastam it up talk about your technology so we're very thankful to know okie and risen them everybody a smart use and we're looking forward to your talk about my shoes yeah thanks Alex thanks