CIBI observability
I instrumented CIBI's Gemini purchase judgments with LangSmith during the cibi-scan-dev development rollout, so failed decisions could be traced instead of guessed at. This is not production tracing.
- React Native
- Gemini
- LangSmith
Software and AI Engineer · Morocco
I work across agent evaluation, LangSmith observability, GIS, and production web products. The case studies here show the traces, tests, and tradeoffs.
Selected engineering work
A real product implementation, a public evaluation system, and GIS work. Each example separates what shipped from what was built to teach.
I instrumented CIBI's Gemini purchase judgments with LangSmith during the cibi-scan-dev development rollout, so failed decisions could be traced instead of guessed at. This is not production tracing.
I built a public multi-agent reference system with invented data and deterministic mocks to reproduce orchestration failures safely. It is not a production deployment.
I engineer responsive geospatial interfaces that stay usable with large GeoJSON datasets and real product constraints.
Case studies

CIBI (Can I Buy It) is an iOS budgeting app that helps salaried professionals make smarter purchase decisions. Scan or enter a price to instantly check it against your budget and savings goals, then get a clear Safe, Careful, or Over verdict before you buy.

Every two weeks, I share one practical note on AI systems, evaluation, observability, or product engineering.