Anirudh User

I built AURA, an agentic multimodal RAG framework for personalised sleep analysis, combining SQLite biometric data, Pinecone journal embeddings, LangChain routing, Gemini, and a Hugging Face cross-encoder reranker. I deployed it end-to-end on Streamlit Cloud and Supabase.

I'm completing an MCA at Vellore Institute of Technology and have built projects spanning fine-tuned BERT classification, sentiment analysis, semantic search, and interpretable computer vision. My DreamWeaver project achieved 93% average accuracy across 12 Jungian archetypes, while my lung cancer model achieved 88% average test accuracy using DenseNet169, CBAM, test-time augmentation, and Grad-CAM.