Selected Work
Backends, platforms and AI tooling — a few projects I'm proud of, pulled from production work and open source. More shipping soon.
Code & databases → Neo4j knowledge graph
Turns Git repositories and relational databases into a Neo4j knowledge graph that you can query over MCP. Maps commits, files, authors and schema relationships into a queryable graph so an agent (or a human) can explore how a codebase actually fits together.
A supervisor for your AI coding agent
Gives AI coding agents a built-in supervisor with four guardrails — security, accessibility, validation and data-loss — plus technical-debt tracking and plain-language traffic-light reports. Works with Claude Code, Cursor and more, so agents ship code you can trust.
Learn AI by thinking, not just using
A thinking-first AI learning platform: 15 tracks and 279 zero-video modules, from "I'm 12 and curious" to a 54-module Zero-to-AI-Engineer path. Built on five pillars — Thinking, Making, Judgment, Expression and Responsibility — with age-based programs for kids, teens and college, plus cross-device progress, streaks and capstones.
The right LLM technique, automatically
Picks the right LLM technique for any task by querying a curated wiki of 20+ agent patterns — RAG, ReAct, Reflexion and more. A rule-based selector plus working runners, routed through llm-mcp with zero API keys.
Self-service edge infrastructure
A production-style internal platform for edge and load-balancing infrastructure: FastAPI control plane, async workers, dynamic Envoy configuration and agentic AI assistance.
Turn any book into a skill
book-to-skill: converts a PDF, Markdown or DOCX book into a Claude Code skill — with a skill mode and a provider-agnostic API mode so the knowledge becomes callable.
Igniting ideas with Python
A growing collection of intermediate Python scripts and snippets crafted to fuel creativity — practical building blocks across data, automation and algorithms.
7 projects · more coming