Scale By The Bay 2018: Sky Thomas Interview
Recording: Scale By The Bay 2018: Sky Thomas Interview
sure um my name is Skylar Thomas I work for map our I am a principal engineer there and I live in San Jose right now we're building a ML and AI environment for kubernetes and the coolest thing about it is about five years ago I left IBM to start a startup that was focused on healthcare I had a younger sister that passed away from heart disease and medical mistakes and I wanted to see at the time whether we could use machine learning and scale big data - and Internet of Things watches with heart rates - to see if we could find a find out when people were have beginning to have health problems and encourage them to go to their doctor at the time it was really hard to build systems that allowed data scientists to experiment and play and and see if there was really really a signal in the data and try lots of different approaches today with kubernetes I think we're at a point in time that that is very possible and that's why I'm really excited about the stuff I'm doing it's it's something I had hoped to build a couple of years ago but we're finally at a point with technology where it's possible I said that's hard there are they're awful really great data pipelines and software architecture are probably the ones that appeal the most so a lot of my talk today was about data scientists are a really unusual breed of user for IT systems data scientists are called scientists for a reason they they actually have to experiment and try different things and cookie cutter software is really hard for them to use they have to be in charge of what they're building at the same time the types of workloads that they're building mean that you don't just want to give every data scientist millions and millions of dollars of hardware to use how he want you'd like to be more efficient that creates a lot of challenges for IT and so some of the best practices that I recommended today we're taking advantage of some of the native features in kubernetes things like namespaces and resource quotas things like kubernetes operators to make make those two worlds fit together a lot better I've seen a lot of talks today on a lot of different areas of kubernetes one that one of the cool things about kubernetes today is that it's growing so fast and there's so many different ways to use things and so many ideas where we're really in the early day is of a new way of doing computing and so going around to a lot of the kubernetes talks here and seeing what everybody is coming up with is really eye-opening and kind of groundbreaking [Music] thank I I haven't I started out doing small talk forever and I have my first experience with Scala was about six or seven years ago when I started doing a lot of AI and ml work and Scala was the language that I know I yeah Scala scholar was a language that allowed me to build workloads that were interesting and I I think that's one of the cool things about scale that's not fair right there they're too many they're really really really too many I don't want to leave somebody out you