• MSL (GenAI)

• Drove ML training framework for the team, supporting models in retrieval and ranking on large clusters (> 1,000 GPUs)

• Model scale up and efficiency tuning for memory footprint and QPS

• AI Infra

• Integrated core training frameworks (Lightning, TorchTNT) to support Ads and Feeds models exceeding 10B parameters, focusing on monitoring, serialization compatibility, and training stability.

Led a 6-person Simulation & Testing team, directing the end-to-end development of high-fidelity virtual environments for autonomous truck testing. Spearheaded the design of a ROS2-based deterministic simulator, significantly increasing the reliability of safety-critical edge case testing

Founding engineer Autonomous vehicle research center for robo-taxi.

Built a distributed system to support large scale simulation tasks (2000+ containers, 100+ jobs per day) scheduling, computation, storage and analysis.

Uber Eats

TL in projects of operation efficiency tool, customer support tools and eater delivery experience

Worked on Skydio’s data, build and testing infrastructures