i build AI systems and like understanding how they work.
some things i've built:
- tribeAI — a minimal agent harness in python. explicit agent loop, context compaction at 60% of the model's budget, and tools confined to a workspace root behind an approval gate.
- EvalWatch — LLM observability platform where apps post traces and a judge model (LLM-As-A-Judge) scores each one for relevance and groundedness.
- UNaFIED — realtime two way chat with an agent always ready to be called. Just say
@unafied schedule a quick call with them @ 5pm. - petique-app — vet appointment platform.
older stuff:
- cliipu — clipboard manager for OMARCHY.
- MemoryWeave — real-time object detections woven into narrative timelines. 1st of 200+ teams at hacksrm'25.
currently:
- AI inference
- LLM internals
- backend systems
- developer tooling
- research
if you want to know more about me, you can head over here to my portfolio!!!!!.
;)