
An autonomous spacecraft crew
Nine ship logs go in. Classification, triage, memory, diagnosis, a safety gate and a ticket come out, with a black box at the end.
- agent crew
- FAISS memory
- safety gates
These are the tools that show up in the work, not everything I have tried.
Firmware, microcontrollers, hardware interfaces
Computer vision, model development, AI
Applications, APIs, infrastructure
Five projects, most of them about systems that have to be right. Agent crews, embedded sensors, and the dashboards that keep them honest.
It is the same thread in every project here: a classifier with a calibrated threshold, an agent that hands back the reasoning behind a decision, a model that reports its confidence instead of guessing.
I work on agentic systems, embedded hardware and the web, usually where those three meet.
Most of what I build is about verification. Classifiers with calibrated thresholds, deterministic safety gates, evidence trails. An agent that cannot show its reasoning is a demo, not a tool.
I am open to internships, freelance work, and problems that need a system rather than a demo. If it involves agents, sensors or evidence trails, it is probably a good fit. Email works best.