Developed a Retrieval-Augmented Generation (RAG) chatbot for Indian budget PDFs. • Supported multilingual input/output using Deep Translator and gTTS for 10+ Indian languages. • Used Pinecone for semantic document search and Gemini API for accurate response generation.

BuChat - A multilingual RAG chatbot for Singapore Budget 2025 using open-source models. Uses Llama 3.2:3b as the ReAct model and a knowledge base query tool, gain observability into the chatbot via Arize Phoenix, Postgres vector store, LiteLLM as the LLM gateway, Mistral OCR API for document parsing, LlamaIndex as the RAG framework.

AI Agent for Digital Financial Literacy A multilingual chatbot powered by IBM Granite and deployed on IBM Cloud Lite. It helps users understand UPI, budgeting, interest rates, and online safety through conversational guidance. Built with Retrieval-Augmented Generation (RAG) for accurate, real-time financial education.