Ray Meetup at Galvanize SF!

Join us for Ray (http://ray.io/) meetup on January 30th at 6:00pm, hosted at Galvanize SF! Engineers at Microsoft, Anyscale, and Numenta will give talks ranging from new improvements in Ray to how they're using Ray to power their work. Be sure to fill out the form below to complete RSVP - we welcome all!

ATTEND

1. RSVP on meetup
2. Fill out this short form: https://forms.gle/E7mu1Y5uQ1AnCJpy5
3. After both steps, we'll have a badge ready for you when you arrive!

AGENDA

6:00pm: Networking, Pizza and drinks!
6:30pm: Ray re-architecture: Fast Scheduling in Ray 0.8 by Edward Oakes.
6:50pm: User Talk: How Numenta is using Ray for Scaling Experiments.
7:10pm: User Talk: How Microsoft is using Ray and RLlib inside a Machine Teaching Service for Autonomous Systems
7:30pm: Ray.Serve: A new scalable machine learning model serving on Ray by Simon Mo
8:00pm: Additional Q&A and Networking

# Talk 1: Ray re-architecture: Fast Scheduling in Ray 0.8

The Ray 0.8 release enables gRPC based direct calls by default, which eliminates the scheduler as a bottleneck and improves performance 5-10x for scheduler-constrained workloads. This talk will discuss the motivation for this rearchitecture and design decisions we’ve made that have enabled this performance.

Speaker: Edward Oakes is a Software Engineer at Anyscale.

# Talk 2: How Numenta is using Ray for Scaling Experiments.

Inspired by the sparsity of the brain, we seek to carry over the benefits of robustness and computational efficiency by training sparse neural networks. This talk will go over, at a high level, why we care about sparsity and some of the different methods we've been exploring to train such networks. All the while, I'll show how we leverage Ray to streamline and scale our experiments.

Speaker: Michaelangelo Caporale is a Software Engineer at Numenta Research.

# Talk 3: How Microsoft is using Ray and RLlib inside a Machine Teaching Service for Autonomous Systems

We foresee a future where libraries for Reinforcement Learning will be well stablished and seamless integrated with Distribute Computing engines. In this brief talk we will show you how we are betting on RLlib and Ray as the framework for Deep Reinforcement Learning that will position us right away in that future that we envision and help us to accelerate our journey of creating a Machine Teaching solution: a complete toolchain for building, training & deploying high-level models that incorporate subject matter expertise from people without CS and ML background.

Speaker: Edi Palencia is a Software Engineer at Microsoft.

# Talk 4: Ray.Serve: A new scalable machine learning model serving on Ray

When a machine learning model needs to served for interactive use cases, the models are either wrapped inside a Flask server or deployed using external services like Sagemaker. Both methods come with flaws. In this talk, you will learn about how ray serve uses ray to address the limitations of current approaches and enable scalable model serving.

Speaker: Simon Mo is a Software Engineer at Anyscale.

****
We look forward to seeing you soon!