Jeshpreet Mahun — Coder & founder working across AI/ML and startup builds — knowledge graphs, multi-agent pipelines, and things that ship at hackathon speed. Currently at OIST.
An AI security layer that protects a company's vector database from prompt injections, malicious inputs, and unsafe LLM interactions. Selected from 108 teams down to the Top 7 pitchers, across three rounds — Pitch the Future, The Marketplace, and Shark Tank — pitching before Vishwa Mohan to take the win as Team The Interceptors, with Divishada Rajput.
Ingests data from Slack, GitHub, Linear, Notion, and Zoom into a causal knowledge graph to detect execution drift and generate specs for AI coding agents. The core long-term startup thesis — currently in active development, first shown at OIST's AI/ML track hackathon with Tanay Pandey.
A conversational crime intelligence platform letting officers query crime databases in English, Kannada, or Kanglish. A router classifies queries into NL2SQL, RAG, graph, or forecasting paths, with confidence-gated responses, source citations, and Neo4j-powered criminal network analysis — built with on-premise LLMs for police-grade security.
A live digital twin dashboard of Bhopal with real-time traffic monitoring, built on React 18, Vite, Tailwind, Framer Motion and React-Leaflet across 10 integrated modules.
A three-agent coordination system — Compliance, Schedule Risk, and Orchestrator — built on Cortex's architecture pattern, targeting AI-driven data centre EPC project delivery.
Two concepts in active development — Chhaya, passive on-device domestic safety audio detection, and Akshar, foundational literacy drift detection for primary school students.