AI and LLM integration
LLM features that work in production: classification, RAG search and tool-calling agents, with evaluation.
How I work on this
I add AI to products where it solves a concrete problem. That includes asynchronous LLM classification behind fast APIs, retrieval-augmented generation over your documents, and customer-operations agents with tool calling, policy guardrails and human-in-the-loop review. Every AI feature ships with an evaluation harness so quality is measured, not assumed.
Technologies
- Python
- FastAPI
- LangChain
- pgvector
- Gemini API
- Langfuse
What is included
- LLM integration (Gemini and other providers)
- RAG pipelines with pgvector or ChromaDB
- AI agents with tool calling and guardrails
- LLM evaluation and observability (Langfuse)
- Background processing so AI never slows the API
Related projects
- Resolve AIEvaluation-driven AI customer operations agent using tool calling, hybrid RAG, policy guardrails and human-in-the-loop workflows.
- DocuMind AIDocument Q&A SaaS built on FastAPI and pgvector RAG, with JWT auth, Stripe billing, async Celery ingestion and Langfuse observability.
- FeedbackIQAI-powered multi-tenant SaaS API that ingests feedback and classifies sentiment, themes and urgency with an LLM in background jobs.
- Production RAG EvalRAG chatbot with an evaluation harness, built with LangChain, ChromaDB, Ollama and Phi-3.
Need help with this?
Let's connect