SOUGATA ROY

I’m pursuing an M.Tech in Computer Science and Engineering with a specialization in Artificial Intelligence and Machine Learning, and I love turning research-grade ideas into systems that users can actually rely on. My background in core CS foundations helps me build solutions that are both efficient and maintainable.

In agentic retrieval, I built an Agentic AI-Powered Multi-Document RAG Assistant that queries multiple PDFs using LLMs and semantic search. I implemented document chunking, embedding-based retrieval, context-aware response generation, and source attribution, and I designed a scalable backend using Node.js, MongoDB, and agent-based routing.

I also explored Deep Learning through a Deepfake Image Detection project, where I designed and trained a custom CNN in PyTorch, applied data augmentation and image preprocessing, and evaluated performance on validation and test datasets for robust detection.

On the application side, I engineered a Full-Stack Financial Dashboard with CRUD capabilities and real-time data visualization using Next.js 14, PostgreSQL, and Prisma. I focused on a type-safe API layer with TypeScript and Prisma ORM, and I resolved deployment challenges in serverless environments by optimizing Prisma engine generation and environment-specific configurations.