Kuldeep Singh

I build production GenAI systems with a focus on RAG & LangGraph-based agents, turning research workflows into reliable, scalable products. Most recently, as a Data Scientist at Fractal Analytics, I architected and deployed a Multi-LLM Reasoning Engine on Databricks using LangGraph—implementing agent orchestration with CoT prompting, intelligent routing, and memory checkpointing that cut query latency by 40% (to 35s), reduced inference cost by 20%, and lowered context-window usage by 60% for large-scale tabular processing.

Before that, at Tata Research & Innovation Labs, I developed a RAG system that automated literature review over thousands of research papers, reducing information retrieval time by 50% through natural-language querying and methodology extraction. I also deployed Vicuna-13b on AWS, achieving a 50% cost reduction versus Azure OpenAI GPT-4 API while maintaining similar chat outputs, and earlier I strengthened my foundations through large-scale data pipeline optimization as an Associate Software Engineer at Impetus Technologies.