Summary
Peng Li is a research scientist and PhD-trained computer scientist based in San Jose with seven years of experience building ML-driven data systems and LLM applications. At ByteDance he led Text-to-SQL research and productization, delivering state-of-the-art results on BIRD and Spider 2.0-Snow and shipping multi-dialect, multi-turn, streaming, and external-knowledge features used across Ads, e-commerce, and analytics teams. His academic work at Georgia Tech produced practical algorithms for learning from dirty data, fuzzy joins without labels, and human-in-the-loop cleaning that bridge data cleaning and downstream ML. Internships at Microsoft and Celonis translated into SOTA table-relationalization methods and KPI-driver discovery, with publications and awards recognized at VLDB and SIGMOD. Known as an “artist working on humanoid intelligence” on GitHub, he blends rigorous theory with product impact and a knack for creative problem framing.
7 years of coding experience