Code & Systems
Open-source focus and systems programming interests.
Most of my day-to-day engineering lives in closed-source systems at Google DeepMind (and, earlier, Lam Research). My public, open-source work lives under ible.ai — a personal venture exploring what is newly possible with language models.
Selected Repositories
Turns any topic into an interactive learning graph: concepts become a navigable web of nodes that stream in live from a local LLM (Ollama), with feedback loops and saveable state. Live demo at ible-ai.github.io/graphible.
Wraps an LLM in a server that learns during idle time — it stores interactions, critiques and revises its own past answers, and fine-tunes on them with LoRA. An experiment in whether some instances "learn to learn" better than others.
Deterministic color-coded debugging for Python: dprint() gives every file-and-line
its own stable color, so you can spot where log output is coming from at a glance. Zero config,
minimal dependencies.