Writing
Long-form notes on systems design, platforms, and the engineering side of scientific work.
Why splitting a research loop into research, adversary, review, and synthesis stages isn't over-engineering, and what it caught when two sibling genes both looked like good features for the wrong reasons.
A retrospective on building a research copilot for feature curation, and why the failure that mattered most wasn't a wrong answer but a confidently wrong one.
I built a fully local meeting recorder in two days with an AI pair. The parts that make it trustworthy are the ones nobody asked for.
A four-year story about turning a research prototype into a versioned, auditable platform, and what changes when you own the production side of a system built where science meets scale.
A draft note on why systems design feels like the bridge between scientific ideas and reliable products.