HackOS 1 - AI Research Hackathon @ UWaterloo x McGill
HackOS 1 is sponsored by Rootly (YC S24)! Learn more about Rootly in the sponsor section below.
Welcome! What is HackOS!?
A student-run hackathon to collectively create models to exceed benchmarks for open access research topics.
Who can join?
This hackathon is geared towards students, researchers, and recent grads. We welcome participants of all skill levels, whether you're writing code for the first time or a seasoned developer.
What is happening?
We'll be hacking together models to beat benchmarks for 4 different streams:
Natural language understanding for LLMs: tinyMMLU (or time permitting, all of tinyBenchmarks!)
Bioinformatics: running predictions on genomic data with Genomic Benchmarks
Vision in compute-limited settings: a subsample of COCO minitrain
WFB: Whatever floats your boat! Choose your own goal/benchmark and hack away! Show us at the end of the hackathon pls <3
How will we evaluate models?
We're planning to create a Python library that you can clone from GitHub to easily run the above benchmarks (except DIY) - still a work in progress, we'll keep you posted!
Tentative Schedule:
Friday, October 26: (6PM to 9PM)
6:00 PM: Doors open
6:30 PM: Intro presentation
7:00 PM: Hacking starts
Saturday, October 27: (10AM to 3PM)
10:00 AM: Start of in-person hacking
12:00 PM: Lunch
2:00 PM: Demos
3:00 PM: Wrap up!
Our friends at McGill are hosting at the same time! We'll chat with them virtually throughout the hackathon.
About Rootly, our sponsor
Rootly (YC S24) is a fast-growing Canadian startup trusted by industry leaders like NVIDIA, LinkedIn, and Wealthsimple to supercharge their on-call and incident response. Our platform seamlessly integrates with your entire tech stack and leverages AI to help SREs and DevOps teams resolve issues faster, reduce downtime, and prevent repeat incidents. Our goal with AI is to fundamentally help companies tackle incidents more efficiently and ensure they don’t happen again.
Questions?
Feel free to reach out to Akira! (email: akirayoshiyama1 [at] gmail [dot] com).
Location
Location: Go to MC 2037 (Velocity Dungeon). dw it's much nicer than the rest of MC.