A hybrid AI + IoT crowd monitoring system for real-time stampede risk detection and emergency response coordination. DSHIELD combines ML inference models with physics-based crowd behavior analysis in a single hybrid backend, deployed on edge hardware and connected to cloud infrastructure for real-time alerting.

Built a closed-loop alert system that triggers automatic notifications to event organizers and dispatches mobile-connected volunteers for coordinated crowd evacuation upon risk threshold breach. The system implements a human-in-the-loop decision layer that inserts manual validation checkpoints before alert escalation, ensuring operational feasibility and reducing false-positive evacuations in real-world deployments.

A full-stack SaaS platform for building trigger-based workflow automations across 7 external platforms — Google Drive, Gmail, Gemini, Slack, Discord, Notion, and Google Calendar — without relying on Zapier or similar no-code middleware. Built with an event-driven runtime, asynchronous execution, and Stripe billing.

Designed event-driven workflow runtime with asynchronous task execution, webhook triggers, and dependency-aware node orchestration. Implemented custom OAuth 2.0 integrations and secure token lifecycle management for all 7 third-party services — handling token refresh, scope validation, and revocation natively. Integrated Stripe subscription and usage-based credit billing, allowing teams to cap workflow execution spend and automatically pause execution when credit thresholds are hit.

A production-grade full-stack social networking platform with graph-based relationship modeling, end-to-end encrypted messaging, and a second-degree recommendation engine. Built for students and alumni to connect, share content, and collaborate through a modern social ecosystem.

Designed a directed-edge graph data model in MongoDB for user relationships, enabling O(1) follower lookups and efficient follow-request state management at scale. Built a graph-based recommendation engine using second-degree connection traversal and mutual-connection scoring. Secured all chat data with AES-256-GCM encryption before database persistence — ensuring messages are unreadable even in the event of a direct database breach. Improved backend throughput by layering Redis-compatible caching and route-level rate limiting to prevent abuse during traffic spikes.

A decentralized exam integrity platform using blockchain technology and cryptographic security to prevent paper leaks and tampering at the infrastructure level. Built in direct response to institutional failures like the NEET paper leak — where centralized systems created single points of compromise.

Built so that paper leaks like NEET cannot happen — and if tampering is attempted, the blockchain identifies exactly where and by whom. Implements Solidity smart contracts for tamper detection, RSA signature verification for identity, and AES-256 encryption for content security.