Where I've worked

Enabling PyTorch operator coverage for Reinforcement Learning (PPO, GRPO, DPO) and SFT workloads on Trainium.

Tracing for coding agents and Vidur simulation for vLLM.

Predictive auto-scaling for High Traffic Events like Prime Day.

Full-stack for the University Team

Hardware testing for NASA's Artemis II and nuclear fission simulations.