Summary
Pei Chen is an Applied Scientist at Amazon with eight years of experience specializing in context-aware LLM post-training and verifiable rubric-based evaluation for agentic systems at scale. She designs and deploys end-to-end post-training data flywheels—live-traffic failure mining, automated root-cause analysis, and difficulty-aware selection—that have been adopted across teams and power 100B+ MoE backbones. Her research spans long-context modeling, RAG, personalization, multi-turn consistency, and GRPO-driven human-alignment, backed by 20+ publications including a NeurIPS Spotlight and a US patent. Comfortable moving models from research to production, she has hands-on expertise in continual pre-training, SFT, DPO, and stack tooling such as Megatron and Verl/Slime. Based in the Bay Area with a Ph.D. from Texas A&M, she also mentors junior researchers and translates rubric-driven evaluation into measurable RL optimization gains.
8 years of coding experience
3 years of employment as a software developer
Master's degree, Finance, General, Master's degree, Finance, General at Southwestern University of Finance and Economics
Ph.D., Computer Science, Ph.D., Computer Science at Texas A&M University
Bachelor of Engineering - BE, Modeling, Virtual Environments and Simulation, Bachelor of Engineering - BE, Modeling, Virtual Environments and Simulation at National University of Defense Technology