Vihan Gaur

I'm building AI-powered systems that combine full-stack engineering, retrieval, and machine learning for real-world problems. My projects span investigative intelligence, secure multi-agent orchestration, trajectory prediction, and connectivity-aware routing.

I built a Crime Analytics & Investigation Assistant using FastAPI, React, TypeScript, LLMs, Vector RAG, Graph RAG, SQL retrieval, knowledge graphs, and SQLite to analyze case records, FIRs, criminal profiles, evidence, and entity relationships. I also developed a Multi-Agent Harness using Python, FastAPI, LangGraph, and NetworkX to monitor agent goals, prevent goal drift, and stop workflows that violate the initial request.

For urban mobility, I've built an intent-aware trajectory prediction system using PyTorch, STGCN, Transformers, ResNet-18, FlashAttention-2, nuScenes, FastAPI, and Docker. I've also created a network-aware routing system that predicts mobile dead zones and travel times to rank connectivity-aware routes.

I'm a winner of IBM Sponsored HackVerse 2.0 and a first runner-up in HARMAN-sponsored MAHE-MOBILITY on the AI track. I'm currently pursuing a B.Tech in Computer Science and Engineering at Manipal Institute of Technology.