I build production-style AI systems focused on RAG, evaluation, agentic workflows, observability, and full-stack delivery.
| Focus | What I build |
|---|---|
| RAG Systems | Grounded retrieval pipelines with measurable quality |
| Agentic AI | Tool-connected workflows with memory, routing, and review |
| Evaluation | Faithfulness, relevancy, context recall, regression checks |
| Observability | Tracing, metrics, structured logs, auditability |
| Delivery | FastAPI, Next.js, AWS, Postgres, Pinecone |
Production-oriented RAG application built with FastAPI, OpenAI, Pinecone, Ragas, Prometheus, Grafana, and AWS.
Key outcomes
0.78faithfulness0.62answer relevancy0.75context recall- deployed on AWS with observability and offline evaluation
What it shows
- grounded answer generation
- retrieval vs generation evaluation separation
- production-style monitoring and benchmarking
Production-style legal review backend built with FastAPI, LangGraph, Postgres, Pinecone, and Phoenix.
Key outcomes
- multi-agent orchestration
- human-in-the-loop approval and revision
- Postgres-backed workflow persistence
- offline evaluation and trace-based debugging
What it shows
- agentic workflow design
- persistent review state
- retrieval + orchestration + observability
RAG & Retrieval ββββββββββ High
LLM Evaluation ββββββββββ Strong
Agentic Workflows ββββββββββ Strong
Observability ββββββββββ Strong
Full-Stack AI Delivery ββββββββββ Strong

