Loading…

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Lightning Talk Sessions [clear filter]
Monday, November 18
 

5:20pm PST

Lightning Talk: Kuber-What-Es?! Misadventures in Building UIs for K8s-Based ML Platforms - Alexandra Johnson, Independent
Last year, our team set out to build a machine learning platform for launching hyperparameter optimization jobs. However, after our launch, the kubernetes-based machine learning platform wasn't a hit with our internal users. This talk explains why, starting with our three critical user interface design decisions, the two problems these led to, and our one direction for future work. This talk is meant to be a very compact, but honest, look at some of the user experience challenges faced by teams building kubernetes-based ML platforms.

Speakers
avatar for Alexandra Johnson

Alexandra Johnson

Tech Lead, Independent
Alexandra loves creating simple and easy to use interfaces for complicated products. Until recently, she was the Platform Tech Lead at SigOpt, where she and her team worked on everything from machine learning infrastructure to web dashboards to API design. She is based out of San... Read More →



Monday November 18, 2019 5:20pm - 5:25pm PST
Exhibit Hall AB - San Diego Convention Center Ground Level

5:32pm PST

Lightning Talk: Sharing a GPU Among Multiple Containers - Patrick McQuighan, Algorithmia
We’ve been sharing GPU resources across multiple containers since 2016, and we’ve learned a few lessons along the way. In this lightning talk, we’ll walk you through the work we’ve done and discuss some newer approaches to the same problem.

Specifically, this talk will address:

* Why GPUs matter
* What makes sharing GPUs across containers hard
* How we’ve managed to share GPUs in the past
* Recent solutions, including the GPUShare Scheduler Extender project (https://github.com/AliyunContainerService/gpushare-scheduler-extender)


Speakers
avatar for Patrick McQuighan

Patrick McQuighan

Senior Software Engineer, Algorithmia
Patrick joined Algorithmia in December 2015 and has focused on improving system performance and creating the Enterprise AI Layer Enterprise product, an ML deployment and management system that runs on multiple cloud providers and on-prem infrastructures. Previously, Patrick worked... Read More →



Monday November 18, 2019 5:32pm - 5:37pm PST
Exhibit Hall AB - San Diego Convention Center Ground Level