Swethanjali Brahma

LLM-Powered Document Intelligence & RAG Pipeline

Built an end-to-end Retrieval-Augmented Generation (RAG) pipeline using LangChain and Groq (Llama 3.3-70B-Versatile) to extract, summarize, and answer questions over complex, unstructured PDF documents. ● Architected a vector ingestion workflow with Unstructured data loaders and CharacterTextSplitter to chunk and preprocess documents with minimal contextual loss. ● Implemented semantic search using HuggingFace Embeddings and a Chroma vector database for precise, similarity-based context retrieval. ● Instrumented production-grade LLM observability with LangSmith for execution tracing, chain evaluation, and latency monitoring, improving debuggability of multi-step chains.