Architecting agentic AI systems and scalable automation platforms deployed to real users.
I'm building the AI Trust Layer — infrastructure that makes AI outputs verifiable, adversarially resilient, and governable.
Most AI builders focus on making models smarter. I focus on making them trustworthy. Over the past year I've shipped three production systems — a hallucination firewall (Top 50 globally, AWS AIdeas), an agentic scam intelligence platform (Top 2% of 40,000+), and a B2B AI governance marketplace — that together form a unified trust infrastructure for AI pipelines.
Think: SSL for AI.
- 🤖 Agentic AI — Built autonomous LLM agents deployed to real users
- 🔧 Full-Stack — Next.js + FastAPI + TypeScript, shipped end-to-end
- 📱 Mobile — 2 apps live on Google Play Store
- 📄 Research — IEEE & i-manager published, Deepfake Detection
- 🏆 Recognition — National winner IIT Delhi · HCL GUVI Top 2% of 40,000+ · AWS AIdeas Top 50 Global Finalist
Frontend & Mobile
Backend & Cloud
AI & LLM
🏗️ AI Trust Layer — Three production systems, one platform
Real-time hallucination detection API — verifies AI outputs against source documents in under 100ms.
- Serverless AWS (Lambda + API Gateway + DynamoDB) · Bedrock Titan Embeddings V2 (1024-dim)
- Claim extraction → cosine similarity → confidence pipeline (VERIFIED / UNCERTAIN / UNSUPPORTED)
- 94.53% verification precision · Python SDK + TypeScript SDK · Next.js real-time dashboard
- Top 50 Global Finalist — AWS 10,000 AIdeas Competition (Global)
The Autonomous AI Operating System for Startups. Uses multi-agent pipelines to automatically analyze ideas and conduct market research.
- Multi-agent LangGraph pipeline · Bi-directional Notion MCP integration
- Automated market analysis, competitor research, and startup execution drafting
- Stack: TypeScript · LangGraph · Notion API · Next.js
Autonomous AI honeypot that engages scammers, extracts financial intelligence, and classifies scam types in real time.
- 5 adaptive LLM personas (LLaMA 3.3-70b via Groq) · 8 scam type classifications · 98% extraction accuracy
- 13+ FastAPI REST endpoints · Next.js 14 dashboard · MongoDB Atlas · Deployed on Render + Vercel
- Grand Finalist — India AI Impact Buildathon 2026 (Top 2% of 40,000+ participants)
Full-stack AI SaaS marketplace combining Product Hunt + G2 + Gartner capabilities for Series A–D startups.
- AI Trust Score, sentiment analysis, upvote system, enterprise pilot marketplace
- Role-based access (Founder, Buyer, Admin) · 50+ REST API endpoints · 7,000+ lines of code
- Solo-built and fully deployed — Next.js 14 · FastAPI · PostgreSQL · Supabase · Clerk Auth
AI-native workspace where natural language generates state-managed UI components in real time.
- 9 generative components · 10 AI tools · GitHub MCP integration · Voice input · Workspace persistence
- Built for Tambo AI "UI Strikes Back" Hackathon 2026
- Stack: Next.js 15 · TypeScript · Tambo AI SDK · Framer Motion · Tailwind CSS
| Paper | Venue | Status | Links |
|---|---|---|---|
| Comprehensive Deepfake Detection: A Comparative Study on Image and Video Forgery | IEEE PuneCon 2025 | Published | IEEE · GitHub |
| Advanced Deepfake Image Detection: A Robust Framework using InceptionV3 and Xception | i-manager's JIP | Published | DOI |
| Award | Event | Details |
|---|---|---|
| 🥇 National Winner (Top 15 of 15,000+ teams) | SBI Youth Ideathon 2025 — IIT Delhi | INR 20,000 Prize |
| 🏅 Grand Finalist — Top 2% of 40,000+ | India AI Impact Buildathon 2026 (HCL GUVI) | Presented at Bharat Mandapam, New Delhi |
| 🌐 Top 50 Global Finalist | AWS 10,000 AIdeas Competition | TruthLayer — AI Hallucination Firewall |
| 🎖️ National Innovation Finalist | Azim Premji University & BITS Pilani Hyderabad | AI-driven Security System |
| 🥈 1st Runner-Up | Hack Wack 2.0 2025 | Deepfake Detection (300+ participants) |
| 🎯 1st Runner-Up | HackAlthon — Azure Developer Community | |
| 📱 2 Apps on Google Play Store | Anishree Technologies Internship | Shipped within 3-month internship |
