scale.bythebay.io: Yuhao Yang Interview

[Music] [Music] so yeah my name is Joo ho-young and I'm so yeah my name is Joo ho-young and I'm from the bigoted happen ology Department from the bigoted happen ology Department of Intel and currently I'm focusing on of Intel and currently I'm focusing on developing large-scale deep learning developing large-scale deep learning applications on spark cluster ok applications on spark cluster ok currently building a library called big currently building a library called big DL and it's for helping users to DL and it's for helping users to developing deeper deep learning developing deeper deep learning applications on spark lancers so we want applications on spark lancers so we want to approve wipped users advise the to approve wipped users advise the capacitance inhale based architecture so capacitance inhale based architecture so we are trying to resolve the issue we are trying to resolve the issue including like large security planning including like large security planning implications and we want to achieve with implications and we want to achieve with a function parity with other like a a function parity with other like a popular deep learning frameworks popular deep learning frameworks tensorflow cafe and others so the goal tensorflow cafe and others so the goal here is really to achieve well to help here is really to achieve well to help user to achieve the most of which big user to achieve the most of which big deal for their deep learning deal for their deep learning applications actually scale circle applications actually scale circle ability is our stress actually for video ability is our stress actually for video and since we built a deep learning and since we built a deep learning applications on spark so it provides applications on spark so it provides some nature scale-out capacity so some nature scale-out capacity so actually that's what we are good at [Music] okay it's a network communication okay it's a network communication actually we need to transfer some actually we need to transfer some greetings and larger models through the greetings and larger models through the networks and your rage can cost us some networks and your rage can cost us some issues for well sometimes the network is issues for well sometimes the network is not stable and sometimes a space is just not stable and sometimes a space is just too slow too slow yes primary issue for us yes primary issue for us [Music] well it has a lot you know we're well it has a lot you know we're primarily using Scala and we found it's primarily using Scala and we found it's quite productive for allocation and for quite productive for allocation and for our customers actually a lot of our customers actually a lot of customers loves to use spark because of customers loves to use spark because of well it's type safe and this efficient well it's type safe and this efficient is possible on Tyrion and the is possible on Tyrion and the programming interface you are they easy programming interface you are they easy to use so yeah we love it a lot [Music] [Music] well it's a quite efficient I can meet a well it's a quite efficient I can meet a lot of fellows in the same industry and lot of fellows in the same industry and learn a lot from other companies and learn a lot from other companies and experice and I can also you know catch experice and I can also you know catch some opportunity for promoting big deal some opportunity for promoting big deal to their deep learning applications [Music]