I started out in marketing, consumer behavior, and product, trying to figure out why people actually do what they do. That dragged me into data science and experimentation to measure it, and eventually into software because I kept wanting to build the things myself.
Right now my focus is in AI engineering/enablement! Building agents or stuff that help me and others build with AI without things breaking in weird ways. I care less about clever prompts and more about whether a system is deterministic, measurable, and genuinely useful.
Fun and growth are the point. I treat learning like progressive overload: pick something slightly outside my comfort zone, break it, figure out why it broke, and add a little more weight to the bar next time. Most repos here started as an itch I wanted to scratch or an idea I wanted to test with running code.
A few recent builds:
- ai-code-tutor-os — Personal AI coding tutor & cognitive harness. Pydantic-AI 2.0 multi-agent backend (A2A delegation + dynamic skills), hierarchical memory, and Tauri desktop shell.
- claudegraph — LangGraph-style state graphs for Claude Code. Turns advisory markdown instructions into strict state machines with enforced routing.
- peer-ai — An AI agent that designs and reviews A/B experiments the way a skeptical senior data scientist would, tested against simulated ground truth.
- go-google-mcp — Model Context Protocol (MCP) server in Go for Google Workspace.
Always up for talking AI, enablement, or weird problems.