I built Quick Dataset, a full-stack platform that turns documents into fine-tuning datasets through a multi-stage LLM pipeline. It supports multiple dataset export formats and uses PostgreSQL for schema design, prompt versioning, and asynchronous generation.
For OpenDoc, I built a document intelligence platform with citation-aware answers and multi-document comparison. Its hierarchical RAG combines LangGraph, Qdrant, BM25 sparse retrieval, and Reciprocal Rank Fusion.
I also built a fingerprint-based blood group classifier using a three-model ensemble, reaching 92% test accuracy with Grad-CAM explanations. I deployed it through FastAPI and Docker, with a React interface.
My accepted IEEE ICISC 2026 paper compares a lightweight CNN with pretrained ResNet18 for Alzheimer’s disease classification, reporting 98.3% accuracy versus 40.41%. I’m studying Artificial Intelligence & Machine Learning at Chaitanya Bharathi Institute of Technology.
