Apply here!

A successful first RLOS Fest

In the (northern hemisphere) summer of 2020, we had the absolute pleasure of collaborating with 6 students from 3 continents on solving real problems in the Vowpal Wabbit ecosystem. The Reinforcement Learning Open Source (RLOS) Fest was organized and run by Microsoft Research. RLOS was a remote program, so we were fortunate that as the world was adjusting to new circumstances, we were able to continue as planned.

There were over 200 applications for the program, and it was an incredibly difficult task to narrow down the many impressive applicants. In the end, we landed on a group of 6 and the amazing work done over this summer was a testament to the incredible talent that was attracted to the program. Thank you to all applicants for your interest! See below to learn more about them and see demos of the great work they did.

Onwards to 2021

We’re extra excited that the program is returning for 2021! To learn more about the program head to the MSR page. The list of available projects for this year can be found here.

Our collaborators for 2020

• Sharad Chitlangia
  • Pushing the limits of VW with Flatbuffers
  • Demo
• Cassandra Marcussen
  • Parallelized parsing
  • Demo
• Harish Kamath
  • Conversion of VowpalWabbit models into ONNX format
  • Demo
• Newton Mwai Kinyanjui
  • Library of contextual bandit estimators
  • Demo
• Mark Rucker
  • COBA: A modern benchmarking package for reproducible contextual bandit research
  • Demo
• Milind Agarwal
  • Contextual bandit data visualization with Jupyter notebooks
  • Demo