Welcome! 🖐️
My research and projects center around agentic reinforcement learning infra, data and recipe.
Created & Maintaining:
🎨 Skill2Env | Reinforcing agents with collective Skills. [Code] [Dataset]
🔳 linex | The minimal out-of-box infra for Terminal Agent RL. [Code]
⭐ Polar | The first open agent RL infra that solves any-harness rollout. [Code] [Paper]
🤖 Gentopia | One of the earliest multi-agent [Framework] & [Platform].
🧠 nanoGPX | Clean collection of modern architectures (RoPE, GQA, RMSNorm, MoE, SSM, etc.) beyond nanoGPT. [Code]
🚀 ReWOO | Token-efficient harness via decoupling reasoning from observation. [Code] [Paper]
Contributing:
🦋 molt | Clean RL infra for research, pytorch FSDP (Nemo AutoModel) + vLLM. [Repo]
🍙 slime | Clean RL infra for scaling, Megatron + SGLang. [Repo]
🧰 NAT | Nvidia nemo agent harness toolkits. [Repo]
🌍 NemoGym | Nvidia nemo RL rollout environments. [Repo]