Hafsah Akram

I build AI agents and RAG systems that run in production. Over the past year I shipped a LangGraph flight search agent where traced errors dropped 86% after fixing the routing logic, a hybrid RAG pipeline that improved retrieval precision ~40% over keyword search, and two MCP servers on Apify including one that turns batches of invoice images into clean structured JSON automatically. My stack: LangGraph, FastMCP, FastAPI, Qdrant, Redis, AWS EC2, Docker. I debug with LangSmith, deploy on EC2, and work async-first. Right now I'm going deep on fine-tuning: building a YouTube video pipeline that generates training data from transcripts and summaries, and working through LLM inference internals like KV cache and quantization. I have 2 years of experience building production LLM systems remotely. I work independently, communicate on GitHub and async channels, and deliver without hand-holding. Open to full-time remote roles in AI engineering, agentic systems, or RAG.