Alex L. Zhang

Hey 👋! The first thing you should do is check out vgbench.com, and then go participate in the AMD x GPU MODE $100K GPU coding competition!

I recently graduated after spending a wonderful 4 years at Princeton University advised by Professor Karthik Narasimhan, Dr. Khanh Nguyen, and Professor Kai Li. There, I received the Phillip Goldman ‘86 Senior Prize in Computer Science, which is awarded to the top undergraduate student in the Princeton CS department.

The most recent research (with 🔗's!) I've led / majorly contributed to (🔀 by ⏰):

Benchmarks 🪑:

• VideoGameBench: environment for evaluating VLMs on popular video games!
• KernelBench: environment for evaluating LLM-generated CUDA / Triton kernels.
• SWE-bench Multimodal : Multi-modal task instances for SWE-bench with the original team.

Models 🌎:

• Neo-1: the first deep learning model for de novo molecular generation + structure prediction.
• Language-guided World Models: world model RL agents that can be conditioned on natural language hints.

MLSys ⚡:

• AMD x GPU MODE $100k MI300X GPU kernel competition: write the fastest kernels on MI300s!
• Project Popcorn 🍿: open source effort to build LLM-generated GPU kernels.
• Fast Triton kernels for Ligo Biosciences' open-source AlphaFold3 (950+ ⭐'s).

Feel free to reach out to talk through my email at [x]@princeton.edu where [x]=alzhang. I’m also very active in the GPU MODE community and co-lead the GPU programming leaderboard!

Random. I got into AI research because I wanted to design intelligent bots for video games when I was younger:

• I first started programming by making Android games in Java.
• I used to make and sell PC games, one of which was mildly successful (~100k+ players). UE4 unfinished example.
• My favorite games of all time are Xenoblade Chronicles and Team Fortress 2.