Enterprise Model Context Protocol server exposing 29 AI-accessible tools and 2 real-time resources, letting an LLM run end-to-end delivery dispatch (orders, drivers, assignments) through natural conversation. Autonomous agentic loop on Gemini 2.0 Flash plans and executes multi-step workflows; an AI assignment engine analyzes 10+ live signals (weather, traffic, vehicle fit) with transparent reasoning. 3-layer multi-tenant auth (SSE proxy, validation middleware, DB-level filtering) with SHA-256 hashed keys; deployed on HuggingFace Spaces via Docker.

MCU-inspired holographic AI personal assistant with task management, AI chat via Gemini, voice input, persistent memory, and 4-way hybrid search. Features multi-agent orchestration with Coder, Researcher, and Browser sub-agents. Built with React 19, Express 5, PostgreSQL + pgvector on Neon DB, and includes an Android app with native push notifications.

Dedicated meme-sharing platform with Google Gemini multimodal AI for automated meme detection and quality scoring at post time, real-time messaging and notifications over SignalR hubs (1:1 and group chat, read receipts), a multi-signal personalized feed (time decay, diversity interleaving, seen-post tracking) and a gamified 8-tier LaughScore ranking. JWT + BCrypt security, Docker Compose deployment. Published evaluation: 45 ms median API latency, <100 ms real-time delivery, 99.8% uptime — the research behind my first-author ECCT 2026 paper.

Fine-tuned a pretrained EfficientNet-B0 CNN to classify real vs. AI-generated images on the OpenFake dataset (~448K images), reaching 91.6% test accuracy, 0.975 ROC-AUC and 0.976 precision with a production training loop (AMP mixed precision, AdamW + OneCycleLR, EMA, label smoothing, discriminative learning rates) on a Tesla T4. Publication-grade evaluation harness with 12+ metrics, ROC/PR curves, calibration and t-SNE, plus robustness testing under JPEG/blur/downscaling and GradCAM saliency explainability.

AI companion app with real-time voice conversations, speech recognition, WebRTC calls, and personalized memory. Features a customizable AI personality with emotional awareness, persistent relationship memory, and natural conversation flow. Built with FastAPI, React, and PostgreSQL.

A comprehensive, enterprise-grade defense operations management platform designed to streamline military resource coordination, optimize operational efficiency, and enhance strategic decision-making through real-time data analytics and secure role-based access control.

A modern, full-stack e-commerce platform built with React and Express.js, featuring secure authentication, shopping cart functionality, comprehensive admin management capabilities, and real-time product management with MongoDB backend.

MERN platform for university club events with a personalized event recommendation engine, PDF certificate generation for participants, JWT authentication with email verification, and role-based club management.

Distributed Facebook/Instagram scraping platform: FastAPI backend with a Celery + Redis task queue, cookie-session management, browser-fingerprint handling, and automatic orphan-task recovery so long-running scrape jobs survive worker crashes. Admin dashboard for job control and data export.

A comprehensive university management mobile application built with Flutter and Firebase, designed to streamline communication and collaboration within educational institutions. Features multi-authentication support, role-based access control, real-time chat, notice board system, study materials management, and interactive polls.

Global push-to-talk dictation daemon for Linux: raw evdev hotkey capture with Groq Whisper speech-to-text, typing transcribed text into any focused window. Works across Wayland and X11.

A comprehensive collection of machine learning projects demonstrating expertise in various ML algorithms and techniques. This portfolio showcases practical implementations of supervised learning, neural networks, clustering, and real-world applications including medical diagnosis, image processing, and predictive modeling.