I'm Shakes. Most of my work sits where data meets climate and energy. The thread is the same across all of it: learn a thing by building a small version of it, then watch where it breaks.
pitchcasts — a 2026 World Cup forecasting model that simulates the tournament hundreds of times, then scores itself against the real results and admits when it gets one wrong.
wipnote — local-first causal lineage and observability for AI-assisted development. It keeps the chain behind each change: the task, the agent, the commit.
A few years building production Python data infrastructure for large residential solar portfolios.
