Get in touch
Production AI & data systems that pay for themselves.
LLM + geolocation pipeline that found end-of-life IoT devices in the field, identified the holding business, and made hardware recovery routine. Self-hosted Llama, zero token spend.
DBSCAN clustering + SageMaker model flagging battery decline before customer-visible failure. Shifted support from reactive to proactive.
Led full analytical migration off MongoDB onto Redshift, with Glue pipelines and a reporting layer the ops team could query without engineering involvement.
Computer-vision + BLE surgical inventory system replacing clipboards and spreadsheets in NHS operating theatres. iOS, FastAPI, YOLOv8.
Two paid weeks. One clear "build this next".
Best for: teams that suspect there's value in their operational data but can't tell where to start. I work hands-on with your data, not from a deck. Most audits convert into a build.
Fixed-price, fixed-scope, in production.
Best for: teams that already know what they want shipped and need it done properly the first time. Weekly written check-in. You own everything on payment.
Embedded part-time in your team.
Best for: the messy middle — past PoC, not ready to make a full-time AI hire. I plug into your stack and your standups. Minimum 4-week commitment.
Production-style pipeline: connect → profile → PII scan → KPI detection → branded HTML/PDF report. FastAPI, pandas, WeasyPrint, Docker, CI.
Computer-vision and BLE surgical inventory system in trial with NHS hospital trusts. iOS app, FastAPI backend, YOLOv8 detector.
Profile and additional repositories. A call-assistant backend and a small evals harness landing here next.