CV

stickman-rl-lab

A reproducible 2D reinforcement-learning lab: an articulated stickman with ten rigid bodies and eight actuated joints learns to reach a target with PPO through an obstacle curriculum, with a desktop console for live training. Independent deterministic evaluation on the full box-and-platform course: 54/80 successes (67.5%). Natural upright walking is not claimed.

Python PyTorch Stable-Baselines3 Gymnasium PyMunk

ai-anomaly-bureau

A browser investigation game that teaches supervised-learning intuition to first-year students, from a cat classified as bread through overfitting, hidden tests, and controlled experiments. The models really train in the browser, and test labels stay hidden until the player runs an audit. Live demo.

React TypeScript Vite Vitest Playwright

dsh-mcp-gateway

Connects ChatGPT Web to a local DSH runtime over OAuth and MCP, so existing tools, skills, and policies can be called in chat without wrapping each extension.

Python MCP OAuth

neural-choreography

A continuous animation of neural-network mechanisms following Dive into Deep Learning: 49 segments, 11 min 21 s at 1080p60, with 64 automated checks covering the underlying mathematics and rendering. Article.

Python Zanim FFmpeg

deformation-aware-retopology

Retopologizes long, tube-like meshes such as arms, fingers, tails, and hoses into quad edge loops along their longest axis, with bend previews and a Blender add-on. Branching shapes such as whole hands and bodies are not yet handled.

Python Blender

funuono

A QQ group bot built on AstrBot and NapCat, with two custom plugins: a join-request verifier that only ever auto-approves, and Ostrakon, reaction-vote moderation. Merged pull requests deploy automatically once CI passes, and a failed deployment rolls back.

Python AstrBot SQLite

concept2motion

Experimental infrastructure for AI-authored mathematical and machine-learning animations: local patches, fast previews, and evidence-based review. A stable improvement in expressive quality has not yet been shown.

labcouncil

A planned open-source platform where a team of AI agents investigates a researcher's ideas and meets with the researcher on a schedule to decide the next experiment. Currently in planning and bounded validation; there is no platform application yet.