At IBM, I build agentic RAG systems for educational and airline documents, improving answer quality, latency, safety, and auditability.
I designed LangGraph workflows with retrieval, correction, routing, query expansion, map-reduce chunk summarisation, Adaptive RAG, and guardrails for large educational PDF collections. I also built a vector-less BM25 retrieval application for regulated aviation documents, reducing vector database dependencies and embedding costs.
At Infosys, I built a RAG-based conversational document assistant for P&G using Hugging Face embeddings, FAISS HNSW search, LangChain, Groq API models, and Streamlit. I also developed outage detection models for LG and automated text summarisation and keyword extraction for British Telecom.
My earlier work spans NLP sentiment analysis, CAE simulation, and telecom engineering. I bring more than four years of data science experience and practical experience taking machine learning systems from data preparation through evaluation and deployment.
