I built Gandalf, an end-to-end datasheet RAG application that combines PDF ingestion, semantic retrieval across prose and tables, and LLM-generated answers grounded in retrieved sources.
For Gandalf, I developed separate processing paths for prose and tables, integrated embeddings, ChromaDB, and the Groq LLM API, and created a RAGAS-based evaluation workflow. I also served the pipeline through FastAPI and deployed a Streamlit prototype on Hugging Face Spaces.
As a Machine Learning Intern at EME Innovation Hub, I applied regression, classification, clustering, and dimensionality-reduction techniques to real-world datasets. My other projects include a student department prediction system, a face-verification system, and a functional tissue-processing prototype.

