Gaokai Zhang is a software engineer and ML researcher pursuing an MS at Carnegie Mellon, with dual BS degrees from UIUC and Zhejiang University and a decade of practical engineering experience. He has driven cutting-edge long-context LLM and agent research at Microsoft Research Asia—co-leading projects like LoongRL (oral at ICLR 2026) and LongRoPE2 (ICML 2025) that push context lengths to 128k+ and even millions of tokens—while also delivering production-ready recommendation LLMs and SFT pipelines. At CMU he builds scalable, automatically verifiable repository-level task synthesis and test-time-training methods to improve code-generation agents, and he is currently applying those agent skills at NVIDIA. His work blends systems-level efficiency (heterogeneous accelerator-aware training) with robustness and evaluation rigor (SORRY-Bench), reflecting a rare mix of applied research and production engineering. Notably, he has a track record of turning novel RL/LLM recipes into reproducible, scalable artifacts that outperform much larger models on long-reasoning benchmarks.
10 years of coding experience
Tungwah Senior High School
Bachelor of Science - BS, Computer Engineering, Bachelor of Science - BS, Computer Engineering at University of Illinois Urbana-Champaign
Master's degree, Master's degree at Carnegie Mellon University
Bachelor of Engineering - BE, Electrical and Computer Engineering, Bachelor of Engineering - BE, Electrical and Computer Engineering at Zhejiang University
Contributions:17 commits, 12 pushes, 1 branch in 1 year 6 months
spring-bootspringbootelkspringboot3thymeleaf
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.