B.E. Electrical and Computer Engineering · Thapar Institute of Engineering & Technology · Expected 2027
I specialize in moving AI from prototype to production — multi-agent orchestration, RAG pipelines, and security tooling for AI-generated code. My work has been deployed on GCP Cloud Run, published on npm, and recognized at national and international levels.
| Award | Scale |
|---|---|
| National Winner — Smart India Hackathon 2025 | India's largest student hackathon · 10,000+ competing teams |
| International Winner — Ulster University Agentic AI Hackathon | Multi-agent legal AI on GCP · beat global university teams |
| National Finalist — Google Cloud Agentic AI Day 2025 | Top 75 of 9,100+ teams · 30-hour agentic AI build sprint |
| Innovation Award — Cybersecurity AI Hackathon - University of Derby | Out of 200+ global teams |
LegalEase — LLM Legal Intelligence Platform
Legal review in India costs more per hour than most people earn per day. Built a 3-agent pipeline (intake → Gemini 2.5 Pro analysis → risk scoring) deployed on GCP Cloud Run. Anyone can upload a contract and get a structured risk breakdown with severity-rated clauses — delivered over WhatsApp + web. Cut manual legal review time by 70%. Sensitive documents stored in end-to-end encrypted Firestore.
Developer Identity Engine — Multi-Agent GitHub Profiler
5 async AI agents — code quality, collaboration, growth arc, pattern fingerprinting, synthesis — run in parallel via asyncio.gather against real source files, commit history, and PR data. Produces archetype classification and an 18-month growth arc in ~30 seconds. Role Fit mode takes a job description alongside a GitHub URL and returns a scored hire/no-hire verdict with specific code evidence and tailored interview questions. Results stream live via Server-Sent Events.
VibeGuard — Security Audit CLI for AI-Generated Code
npx @vidhaankhare/vibeguard scan .
Most AI-generated apps ship with zero security review. Built a CLI scanner that audits any codebase in one command across 10 layers: prompt injection, slopsquatting, mass assignment, BOLA/IDOR via cross-file taint analysis, live credential verification against full git history, and supply-chain threats generic tools miss. Returns a 0–100 security score with severity-rated findings, exact vulnerable code paths, and copy-paste fixes exportable as JSON/Markdown. No setup — one command from install to full report.
