A collection of systems, security tools, and research projects — from bare-metal OS work to distributed cryptography.
A distributed password vault implementing Shamir's Secret Sharing with Google OAuth and TOTP 2FA. Vault shards are split cryptographically and distributed — no single point holds the secret. React frontend, Flask backend with AES-256.
A zero-trust behavioral biometrics engine that moves beyond cookies and browser fingerprints to biological-physics security. Captures neuromuscular jitter and keystroke dwell/flight times — a 1D-CNN running entirely in-browser computes a humanity score (θ). A DQN agent then dynamically tunes Argon2id hashing strength per login. Bots scoring θ < 0.1 get silently redirected into a honeypot for threat intelligence. A deep autoencoder continuously anchors session trust and flags behavioral drift.
Diploma capstone — a Django web app that detects phishing and malicious URLs using machine learning. Combines Random Forest and Logistic Regression models trained on real phishing datasets, extracting features from URL structure, domain reputation, and metadata. Goes beyond static blacklists with an AI-powered composite scoring engine.
A next-generation contactless attendance platform for institutions. YOLOv8 handles real-time face detection; FaceNet embeddings with FAISS similarity search drive recognition. Liveness detection blocks proxy attendance. A React + TypeScript dashboard gives live check-in status and analytics, while a Flutter mobile app covers student and teacher workflows — all backed by Firebase Realtime Database via async FastAPI endpoints.
3rd-semester EL project diving into OS internals — process scheduling, memory management primitives, and system call interfaces. Built to understand what actually happens beneath the abstraction layers most developers never touch.
Data Structures & Algorithms EL — an interactive platform for visualizing sorting, graph traversal, and dynamic programming algorithms. Bridges the gap between theoretical complexity analysis and observable execution.
An AI-powered adaptive learning platform revolutionizing coding education. Uses LLM agents (GPT-4o / Claude) to deliver personalized tutorials, real-time code review with actionable feedback, and intelligent hint scaffolding—showing just enough guidance without spoiling solutions. Gamified progress tracking with adaptive difficulty keeps learners engaged and motivated.
A collaborative platform designed to help students discover teammates for hackathons, coding competitions, and tech events. Features skill-based team discovery, real-time team chat, direct messaging, and event exploration. Users create profiles with skills and social links, browse events, form or join teams with specific requirements, and communicate seamlessly — all built with a mobile-responsive design powered by Next.js and Supabase.
A smart waste management dashboard built with Flask and modern CSS. Demonstrates how IoT and digital systems can optimize waste segregation in schools, colleges, and smart cities. Features an alarm system that triggers alerts for full bins or hazardous waste detection, a clean Google-inspired dark-mode UI, and is designed for easy extension with real-time sensor data and reporting capabilities.
AI-powered microgrid platform predicting renewable energy with 87% accuracy using LSTM + Transformer ensemble. DQN agents optimize real-time dispatch of loads, battery storage, and grid selling. Reduces peak demand by 18-24%, increases renewable integration through federated learning for collaborative multi-microgrid training without data sharing.