Diganta Pan

I’m an AI backend developer and Computer Science Engineering graduate with hands-on experience designing and deploying end-to-end AI pipelines spanning computer vision, large language models, and retrieval-augmented generation (RAG).

In my AI Backend Developer Intern role, I engineered an event logging and feature pipeline that captured 15+ structured event types across SQL tables, reducing ad-hoc data retrieval time by ~35%. I also improved ranking performance by 23% over a distance-based baseline using an agent scoring model, and built a fraud detection framework with 91% precision across 100+ scenarios.

My project work reflects my focus on real-time, production-ready outcomes. For Project Drishti (Google Cloud Agentic AI Day 2025 finalist), I architected a centralized monitoring system integrating computer vision and LLM summarization, achieving sub-2-second alert latency and 87% relevance on a manual evaluation using Gemini 2.5 Pro API and Vertex AI Agent Builder.

I enjoy taking solutions from offline pipelines to interactive systems—like compressing 60-minute video into 3–5 minute summaries (~15× compression) with Whisper and T5 on CPU-only execution, and building a RAG API chatbot over Upwork reference PDFs with ChromaDB and hallucination-guarded prompting. I bring strong Python, deep learning, and scalable system design to every step.