Projects

12 projects

Policy Surgery

—

Can GPT-OSS-120B edit a LunarLander policy to handle changed flight conditions?

·

Statecraft

—

I built worlds where using a tool changes what the next tool call will do.

·

Sparse Signal Loop

—

I tested what agents should remember when they get very little feedback.

·

Weigou

—

I experimented with training LLaMA-style models across GPUs.

·

Yuntun

—

I trained a Qwen3-style model on FineWeb using tensor parallelism.

·

GPT-2 Speedrun

—

Multi-GPU GPT-2 training with streamed FineWeb-Edu and resumable checkpoints.

·

BeaconGrad

—

I wrote autograd in NumPy and checked its gradients against finite differences and PyTorch.

·

Concrete Score Matching

—

I used score matching to learn from discrete data and Metropolis–Hastings to draw samples.

·