Built a production-grade enterprise AI copilot for unified workplace search using RAG architecture, serving 5K+ daily users and improving knowledge discovery efficiency. Architected an enterprise knowledge platform integrating fragmented tools using Python, LangChain, and LlamaIndex, processing 12M+ documents. Engineered high-performance RAG pipelines using Pinecone and FAISS with OpenAI GPT and LLaMA, enabling permission-aware semantic search. Optimized LLM performance using prompt engineering, context enrichment, and query rewriting, reducing hallucinations and improving response accuracy. Deployed real-time AI search services using FastAPI, Docker, and Kubernetes on AWS, achieving sub-120ms latency. Implemented enterprise LLMOps pipelines using LangSmith, MLflow, and Weights & Biases with Guardrails AI for compliance and security

Built an end-to-end churn prediction platform evolving from analytics to real-time ML system with MLOps and cloud deployment. Improved model ROC-AUC from 0.72 to 0.89 through feature engineering and model optimization. Developed scalable data pipelines integrating PostgreSQL, MySQL, and AWS S3 processing 10M+ records daily. Built machine learning models (Logistic Regression, Random Forest, XGBoost) improving prediction precision by 20%. Deployed real-time inference services using FastAPI, Flask, Docker, and Kubernetes with CI/CD pipelines. Designed executive dashboards in Power BI and Tableau reducing churn by 15% and protecting $1.8M revenue. Conducted exploratory data analysis to identify behavioral trends and enable targeted retention campaigns