Built a Perplexity-style AI research assistant (DOCASK) focused on retrieval quality and reducing hallucination—not just LLM calls.
🔗 https://lnkd.in/gMwY2SuD
Developed an end-to-end RAG system using Python, FastAPI, LangChain, FAISS, and Transformers.
Improved latency (30–40%), relevance (25%), and reduced hallucinations (18%) with hybrid search, semantic chunking, and agentic pipelines.
Open to AI/ML/backend roles.





