Aditya User

I build end-to-end AI applications that connect LLM-powered intelligence to production-grade backends and user interfaces. I’m especially focused on retrieval quality—hybrid search, reranking, and relevance grading—to make AI outputs trustworthy and fast.

In my AI/ML work, I designed a 35,000-edge knowledge graph from 8,000+ business documents using LLM-powered entity extraction and FAISS embeddings for intelligent recommendations. I also architected a hybrid retrieval engine (BM25 + FAISS) with cross-encoder reranking and automated relevance grading, delivering high-precision results across 2,000+ regulatory licenses. I deployed scalable infrastructure with FastAPI, Docker, PostgreSQL, Neo4j, and LangChain.

I bring the same production mindset to workflow automation and full stack delivery. I automated an end-to-end publishing pipeline with n8n and the OpenAI API, cutting WordPress publishing time by 60%, and built a high-performance Python backend with JWT authentication, async REST APIs, and multithreading—improving throughput by 40% for 50+ concurrent users. In projects like Intervue and a medical RAG bot, I ship robust APIs and interfaces using FastAPI, Next.js/React, pgvector, and FAISS—prioritizing latency (under 2s) and efficient, reliable system behavior.