Short write-ups on the things I build: agents, models, signals and the tooling around them. Each one is written up in full, with the diagrams and the demo.
01
From one phone photo to a shareable 3D model: how I built ObjectShare3D
One photo in, an interactive 3D model out. YOLO11 finds the object, SAM 2 cuts it out, TRELLIS.2 reconstructs it with Hunyuan3D-2 as a fallback, and a Three.js viewer shares it. The full implementation, stage by stage.
Seeing without a camera: human activity from WiFi signals
Detecting sitting and standing with no camera or wearable. WiFi Channel State Information from an ESP32 changes when a person moves, so sliding windows, statistical features and a Random Forest can read it in real time.
Seven agents, one animated video: from a topic to the final cut
A MetaGPT-inspired pipeline of seven agents turns a topic into a narrated Manim animation: planner, script writer, code generator, validator, voice generator, composer and merger. AST and execution checks catch broken code before it is rendered.