Zhengyao Jiang
Bio
Zhengyao Jiang is the co-founder and CEO of Weco AI. His research focuses on using LLM-driven agents to automate machine learning research and code optimization.
Session (1)
Zhengyao Jiang is the co-founder and CEO of Weco AI, the team behind AIDE — a tree-search-based ML engineering agent that autonomously writes and improves code to automate AI R&D. His session offers rare insight from someone actively building and shipping evaluation-driven autonomous research agents, with practical lessons on benchmarking, reward hacking, and what it takes for AI to do AI R&D.
GitHub
@ZhengyaoJiang- ZhengyaoJiang/PGPortfolio★ 1,851Policy Gradient Portfolio: deep RL framework for financial portfolio management
- ZhengyaoJiang/latentplan★ 113Efficient Planning in a Compact Latent Action Space (ICLR 2023)
- ZhengyaoJiang/NLRL★ 77Source code of Neural Logic Reinforcement Learning
- WecoAI/aidemlorg★ 1,328AIDE: AI-Driven Exploration in the Space of Code — the ML engineering agent that automates AI R&D
Recent writing (4)
- SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents · blog · May 2026
- AutoResearch vs Classical Hyperparameter Tuning · blog · Apr 2026
- How to Build a Good Eval · blog · Feb 2026
- Weco Raises $8M to Build Self-Evolving Software · blog · Jul 2025
Public activity researched automatically · as of Jun 2026