At UptoSkills, I architected a Generative AI interview system using LangChain, FAISS vector search, and the Groq LLM API. It combined short- and long-term memory with an LLM-as-Judge contradiction checker to evaluate response quality.
I also tuned a Faster-Whisper speech-to-text pipeline on a 150-clip spontaneous-speech dataset, reducing relative Word Error Rate from 0.382 to 0.2753.
For KaiNote, I designed a multi-agent workspace with LangGraph state machines and built a FastAPI backend. ChromaDB retrieval supports contextual, embedding-based memory across notes, tasks, and scheduling.
I’m a final-year B.E. Computer Engineering (AI & ML) student at A.P. Shah Institute of Technology, Thane. My projects also include an AI medical scribe with real-time transcription, structured clinical EHR extraction, and automated prescription PDF generation.

