APURBA BARMAN

I’m an M.Tech. student in Geoinformatics at IIT Bombay, focused on turning complex data into dependable AI and geospatial solutions. I enjoy building from first principles—whether that’s training deep learning models or engineering production-ready pipelines. I’m motivated by systems that improve speed, accuracy, and privacy while staying usable for real teams.

In my current Product Management Intern role at MediaGuru, I led the end-to-end development of an AI-powered Document Intelligence Platform for automated summarization and knowledge extraction. I architected and deployed a production-grade RAG system using LangChain, pgvector, PostgreSQL, FastAPI, and Docker, enabling context-aware retrieval and ~13-second end-to-end summarization. I also evaluated multiple LLM providers to balance cost, latency, and data privacy, documenting APIs with Swagger/OpenAPI for organization-wide adoption.

As a Data Scientist at Makr Microsystems, I build Python-based AFM image preprocessing pipelines (plane leveling, row alignment, FFT noise filtering) and develop automated feature extraction workflows for nanoscale semiconductor structures. I’m exploring deep learning approaches for feature detection and uncertainty estimation in AFM metrology. My projects—from a C++ deep learning framework to an MCP-compliant Expense Tracker server and geospatial clustering/tracking apps—reflect my preference for practical, measurable outcomes and clean, modular implementations.