Founder and engineer, working mostly on the AI infrastructure other people build on top of.
I spent the better part of a decade writing enterprise software before drifting, gradually and then all at once, into machine learning. These days I split my time between building Mightymonk, a company I co-founded, and doing research that keeps me honest about what I actually understand versus what I've just gotten used to.
AI infrastructure and developer tools — evaluation, reliability, and a couple of other things I'm not quite ready to talk about yet.
A research-first company where I get to work further out — edge AI, on-device intelligence, and problems too early to belong in a product roadmap yet.
Not a resume — just a running record of what I've been building or learning, roughly in order.
Started Cognovision Solutions, a company for research further out than a typical product roadmap — on-device intelligence, edge AI, that kind of thing.
Co-founded Mightymonk. Most days this means building evaluation and reliability tooling for LLM-based systems, and occasionally remembering to also run a company.
Started building an uncertainty-aware forecasting model — LSTMs plus conformal prediction — mainly to test whether I actually understood calibration or had just gotten comfortable with the term. Still at it, with more research and improvements in progress; aiming to eventually publish and take it into production. Code's public on
Designed and built an enterprise semantic search platform, stitching together half a dozen internal databases into something people could actually search.
Spent this stretch building enterprise software for research institutions and procurement organizations in the US — semantic search, e-commerce backends, AWS Lambda pipelines, Salesforce integrations, payment gateways, and more server and cloud security firefighting than I'd like to admit.
Wrote a lot of Ember.js for travel software and learned more about shipping under deadline pressure than any course would've taught me.