VLA / VLM robotics builder working on robot brain architecture, simulation data, Real2Sim pipelines, and embodied AI runtime systems.
机器人大小脑 · 多模态感知 · 仿真数据 · Real2Sim · 可观测运行时
Website · Hugging Face · X · 中文 · Français · Русский · عربي · 日本語 · Português · Türkçe
robot-runtime.console
$ boot --stack vla-vlm --mode embodied
> perception=vlm policy=vla sim=real2sim
> brain=planner+controller data=observable
> status=online latency=adaptive loop=closedI build the stack behind robot intelligence: multimodal perception, VLA policy learning, robot big brain / small brain architecture, simulation data engines, Real2Sim assets, and runtime infrastructure that makes behavior inspectable.
flowchart LR
A[World Data] --> B[Simulation Engine]
B --> C[Embodied Dataset]
C --> D[VLM World Model]
D --> E[VLA Policy]
E --> F[Robot Big Brain]
F --> G[Small Brain Runtime]
G --> H[Real Robot Feedback]
H --> A
style A fill:#020617,stroke:#22d3ee,color:#ffffff
style B fill:#020617,stroke:#8b5cf6,color:#ffffff
style C fill:#020617,stroke:#22c55e,color:#ffffff
style D fill:#020617,stroke:#f59e0b,color:#ffffff
style E fill:#020617,stroke:#fb7185,color:#ffffff
style F fill:#020617,stroke:#38bdf8,color:#ffffff
style G fill:#020617,stroke:#a3e635,color:#ffffff
style H fill:#020617,stroke:#facc15,color:#ffffff
| Layer | Direction |
|---|---|
| Robot big brain | multimodal planning, instruction grounding, memory, tool use |
| Robot small brain | motion/runtime orchestration, controller adapters, execution feedback |
| VLA / VLM | scene semantics, action grounding, policy learning, evaluation |
| Simulation data | synthetic scenes, Real2Sim assets, domain randomization, dataset QA |
| AI infrastructure | agents, code automation, model routing, workflow verification |
My public repositories include robotics-adjacent AI infrastructure, developer tools, model routing, code review automation, knowledge workflows, and simulation/product systems. I care about systems that can be observed, debugged, reproduced, and improved instead of only looking impressive in a demo.
| Area | Project | What it does |
|---|---|---|
| AI agents | kakashi | Codex-powered system for searching GitHub capabilities, planning repository fusion, executing changes, and verifying the result. |
| AI tooling | ai_code_reviewer | LLM-based code review automation for GitHub, GitLab, and Gitea, with multi-model support. |
| Model routing | openai-chat-switch | Go package for chat embeddings and model/chat switching workflows. |
| Learning systems | little_language_model | Small language-model experiments and implementation notes. |
| Developer tools | esh | Cross-platform SSH connection manager with encrypted credentials and cluster command execution. |
| Infrastructure | qcow2file | Builds qcow2 VM images from Dockerfile-like recipes. |
| Knowledge workflow | obsidian-image-auto-upload | Obsidian plugin for automatically uploading pasted or dropped images to external storage. |
VLA / VLM / Robotics / Simulation Data / Embodied AI