Harsh Dubey

About Harsh Dubey

I’m a third-year B.Tech IT student at Netaji Subhas University of Technology (NSUT) with a strong foundation in core CS subjects and a passion for backend systems, Generative AI, and scalable web applications.I’ve interned at Sopra Steria, where I built a GenAI-powered RAG pipeline using LangChain, Hugging Face, and Milvus for intelligent document search. At Americana Foods UAE, I developed a fraud detection system by linking POS data with surveillance video using Python, OpenCV, and MySQL. I also worked at Sovratec, where I built a Flask-based admin portal with custom RBAC, session handling, and real-time dashboards.Outside internships, I’ve built projects like a WebRTC-based live streaming app and a Rust-powered terminal chat platform. I enjoy solving real-world problems at the intersection of AI, backend development, and system design.

Experience

1. Software Developer

   Sopra Steria

   Jun 2025 — Present · Noida, IN

   Developed a GenAI-powered Retrieval-Augmented Generation pipeline using LangChain, enabling semantic Q&A over 100+ files. Leveraged sentence-transformers with BAAI/bge-base-en-v1.5 from Hugging Face to create embeddings for 12K+ chunks and query. Stored structured content in MongoDB and embeddings in Zilliz (Milvus Cloud) to enable sub-second top-k similarity search. Integrated a lightweight SLM (FLAN-T5) via Hugging Face Pipeline, to generate context-aware answers grounded in retrieved content. Technologies Used: GenAI, Python, LangChain, Hugging Face Transformers, MongoDB, Zilliz (Milvus), sentence-transformers, SLMs

Education

• Netaji Subhas University of Technology

  Bachelor of Technology - BTech, Information Technology

  2022 — 2026

• Gyan Bharati School - India

  High School Diploma, Computer Science

  2007 — 2022