By day, I architect high-throughput AI compute pipelines and optimize complex models across decentralized networks of heterogeneous GPUs. On the open-source side, I'm a core contributor building stateful distributed backplanes and massive-scale RL infrastructure.
My work bridges research and low-level systems;whether that means scaling distributed RL (GRPO/PPO), eliminating I/O bottlenecks in training clusters, or compiling multimodal backbones into TensorRT for continuous inference.
When I’m not scaling clusters or compiling engines, you’ll find me listening to Shoegaze, tracking astronomy, playing badminton, or leaning into my Otaku side.
Thanks for reading!
-
You can check out some sections of my work through the repos.
-
If you are interested to collaborate on some cool things like reinforcement learning or robotics, feel free to hit me up via email