← All conversations
People & ideas

Scale By The Bay 2019: Oscar Boykin Interview

Oscar Boykin ↗With Alexy Khrabrov20195:03

FunctionalTV interview with Oscar Boykin.

Listen

Download MP3 ↓

Follow the words

Read the transcript

[Music] my name's Oscar Boykin I work at stripe my name's Oscar Boykin I work at stripe I'm a machine learning infrastructure I'm a machine learning infrastructure engineer [Music] this is not my first time speaking in this is not my first time speaking in Scala by the bay scale by the bay you Scala by the bay scale by the bay you can tell by just my slip I've been doing can tell by just my slip I've been doing it for a few years it used to be go it for a few years it used to be go under a different name I first spoke a under a different name I first spoke a few years ago when I was working at few years ago when I was working at Twitter Twitter had it has a close Twitter Twitter had it has a close relationship because of Scala interests relationship because of Scala interests and a lot of that is a big overlap with and a lot of that is a big overlap with the community that's still here my the community that's still here my favorite part of the talk is just kind favorite part of the talk is just kind of like I think we're seeing early days of like I think we're seeing early days of how machine learning is going to be of how machine learning is going to be you know run how it's going to be scored you know run how it's going to be scored how are people going to like use this how are people going to like use this and build it into their production and build it into their production systems so I think people who kind of my systems so I think people who kind of my talk will be able to learn a little bit talk will be able to learn a little bit about a new format called onyx which is about a new format called onyx which is an intermediate representation for an intermediate representation for different machine learning models we're different machine learning models we're not really there yet of a world of like not really there yet of a world of like Java class files or whatever for machine Java class files or whatever for machine learning but onyx is kind of like trying learning but onyx is kind of like trying to take us a little bit in that to take us a little bit in that direction so it'll be interesting to see direction so it'll be interesting to see in the next couple of years if onyx is in the next couple of years if onyx is going to be a standard that really going to be a standard that really catches on but in any case people can catches on but in any case people can learn about like what it's all about yeah so the fact of the matter is even yeah so the fact of the matter is even though a lot of people here at stripe though a lot of people here at stripe like sorry a lot of people here at the like sorry a lot of people here at the scale by the bay like like Scala from scale by the bay like like Scala from machine learning Python is where machine learning Python is where everyone's using Python for the most everyone's using Python for the most part to train their machine learning part to train their machine learning models but on the server side at stripe models but on the server side at stripe we don't really have any Python services we don't really have any Python services so we were really struggling with this so we were really struggling with this question of how can we get deep learning question of how can we get deep learning models out there how can we give more models out there how can we give more more models for more applications and more models for more applications and continue to use our existing Scala continue to use our existing Scala infrastructure so my work will talk infrastructure so my work will talk about a library that we wrote that about a library that we wrote that allowed us to take these models or train allowed us to take these models or train in Python but would score them and run in Python but would score them and run them on our existing java virtual them on our existing java virtual machine based written in scala systems machine based written in scala systems [Music] yeah well Brian Cantrell's talk was yeah well Brian Cantrell's talk was really amazing on just you know Moore's really amazing on just you know Moore's law rights law amazing talk everyone law rights law amazing talk everyone worse worth the price of admission to worse worth the price of admission to scale by the bay alone so I highly scale by the bay alone so I highly recommend everyone take a look at that recommend everyone take a look at that one oh but there are so many really good one oh but there are so many really good talks it's hard to pick yeah that's talks it's hard to pick yeah that's right pick a few it's a great conference right pick a few it's a great conference always good well I'm pretty I don't know if it's well I'm pretty I don't know if it's gonna be over the next year I like a you gonna be over the next year I like a you know been interested in functional know been interested in functional programming for a long time but we're programming for a long time but we're getting closer to the next release of getting closer to the next release of Scala Scala 3 coming out and it has a Scala Scala 3 coming out and it has a lot of really exciting features you know lot of really exciting features you know improves the language in a lot of ways improves the language in a lot of ways simplifies it in many ways but also adds simplifies it in many ways but also adds some nice power add some opportunities some nice power add some opportunities for writing very fast code some for writing very fast code some optimization so I'm pretty excited about optimization so I'm pretty excited about that being something that more people that being something that more people can try to play with and hopefully that can try to play with and hopefully that we can you know eventually start we can you know eventually start including in our work my favorite part I mean it's uh I think my favorite part I mean it's uh I think it's well it's kind of awkward too it's well it's kind of awkward too because scale by the way used to be because scale by the way used to be Scala by the bay and still I would Scala by the bay and still I would imagine more than two-thirds of the imagine more than two-thirds of the attendees have some interest in Scala or attendees have some interest in Scala or domains where Scala has been you know domains where Scala has been you know successful data machine learning things successful data machine learning things like this so there is this ambition for like this so there is this ambition for it to be you know about scalability but it to be you know about scalability but I think it's also still very much I think it's also still very much related to Scala so I like coming and related to Scala so I like coming and saying I think it's really the best saying I think it's really the best Scala conference on the west coast I Scala conference on the west coast I really like coming to see a lot of the really like coming to see a lot of the people Twitter is in in San Francisco people Twitter is in in San Francisco and so they use a lot of scholarly I get and so they use a lot of scholarly I get I used to work at Twitter I get to see a I used to work at Twitter I get to see a lot of my friends who were Twitter lot of my friends who were Twitter engineers there's a lot of them come so engineers there's a lot of them come so it's a great chance to connect with the it's a great chance to connect with the community great venues usually almost community great venues usually almost always a great event always a great event [Music]

Recovered English captions. Automatic transcription may contain errors.

Keep exploring

Follow the guest, their work, and the ideas behind this conversation in the Devreal knowledge graph.

Oscar Boykin on Devreal ↗
Independent by design

Your player.
Your subscription.

One permanent feed. Listen in the podcast app you love, with the conversations always at home here.

https://struct.fm/feed.xml
113 audio episodes available in the feed.