Interests:
- reinforcement learning
- kernel optimization
- distributed training & inference
- Working on RLHF/GDPO pipelines and reward modeling
- Exploring RLMs
Key Achievements:
- Optimized zerobrew package manager 3.6x faster (6min → 1.7min) via racing cancellation and HTTP/2 tuning
- Achieved 12.804s in NVIDIA FP4 GEMM kernel optimization hackathon on B200 using CuTe DSL
- Core contributor to PrimeIntellect-ai Environments-Hub - built 6 evaluation environments for web agents, tool-use, evals
Open to collaborations.