Details
Perch Insights
Fullstack AI Engineer
Worked on natural-language analytics, AI pipeline reliability, evals, observability, and service boundaries for a B2B analytics product.
Live 3D architecture
The scene beside this panel is this system's architecture. Drag to rotate, scroll panel to read on.
What this role proves
Worked on natural-language analytics, AI pipeline reliability, evals, observability, and service boundaries for a B2B analytics product.
Improved the natural-language query pipeline for time, business context, metrics, dimensions, filters, sorting, and customer-specific datasets.
Built and iterated an evaluation framework that helped improve agent precision and reliability by about 30%.
Migrated pipeline observability to LangChain and LangSmith, exposing step-level inputs, outputs, and context for substantially faster debugging and iteration.
Built BFF/proxy/agent-service layers, including a Node.js SSE BFF associated with about 30% lower perceived latency and FastAPI services for heavy analytics work.
Stack and domains
- Python
- FastAPI
- Node.js
- TypeScript
- LangChain
- LangSmith
- PostgreSQL
- PgVector
- SSE