Shib Kumar Saraf

Built a secure Agentic AI Orchestration Platform using FastAPI with JWT authentication, enabling safe multi-tenant enterprise data access. Engineered a hybrid reasoning pipeline combining NL2SQL and vector search to answer natural language queries across databases and documents with explainable outputs.

Developed a full-stack MERN application integrating Google Generative AI to automatically analyze code for quality, security vulnerabilities, performance issues, and best practices. Designed REST APIs for AI-driven reviews and built a real-time React-based code editor with syntax highlighting and structured Markdown feedback.

Built an LLM-driven NLP system leveraging transformer models (BERT, DistilBERT, XLNet, GPT, Gemini), zero-shot classification, and topic modeling (LDA, NMF) to detect mental health signals from high-volume social media streams. Designed scalable data pipelines for text preprocessing, embedding generation, sentiment analysis, and hybrid rule-based modeling to uncover behavioral patterns and emerging mental health trends.

Designed a hybrid recommendation engine using Collaborative Filtering(based on genre, artist, lyrics, etc.), Alternating Least Squares(ALS), and Graph-based similarity models to personalize music discovery. Improved recommendation relevance by integrating lyrical embeddings and audio feature vectors, deployed via an interactive Streamlit interface.

A lightweight, production-ready framework for running Large Language Models (LLMs) locally with a modern web-based chat interface. It enables GPU and CPU execution, works efficiently on systems with as little as 8GB RAM, and delivers response speeds comparable to cloud-hosted chatbots — while keeping all data fully private. Designed for developers, AI engineers, researchers, and privacy-focused deployments.

An AI-powered product for C-DAC providing real-time vigilance and video analysis to identify anomalies and ensure integrity in public exams.