Sourav Roy

I’m an AI/ML Engineer with 3+ years of experience building production-grade LLM systems, including RAG pipelines and agentic AI workflows. I focus on the full LLM stack—from retrieval and ranking to generation and deployment—so models perform reliably in real products.

I’ve shipped low-latency GenAI APIs using FastAPI, serving 10K+ daily requests and delivering 200–300ms end-to-end responses with major latency reductions. I also fine-tune large models with QLoRA, optimize vector retrieval and re-ranking with FAISS, and improve production quality through prompt engineering and context management to reduce hallucination rates. Beyond LLM platforms, I’ve built NLP automation pipelines processing 20,000+ records/day and developed backend APIs handling 15,000+ daily requests with meaningful inference-time improvements.