Raghib
Haque

A WarioWare-style micro-game engine where all game logic lives inside a custom Linux kernel module. The kernel exposes /dev/kernelware as a character device — userspace reads events, writes inputs, and calls ioctls to drive 7 mini-games. Built with C using hrtimers, kthreads, spinlocks, waitqueues, and ncurses. Features real hostname manipulation (init_uts_ns), live kmalloc/kfree memory management, /proc/kernelware filesystem stats, and an in-kernel leaderboard. Co-authored with Aidan, Cathal & Tom for Dr. Mark Burkley's course.

A collaborative project built in Java. A practical exercise in version control workflows and collaborative development using Git. Features robust switch-case logic and proper project structure with Maven build management (pom.xml). Demonstrates clean code practices, team collaboration workflows, and practical software engineering fundamentals.

A JavaFX text-adventure game built for CS4431, following a full MVC architecture. Features a multi-room world, NPC dialogue system, puzzle mechanics with item-based unlocking, and a minimap rendered live in the GUI. Implements core Java concepts including Serialization for save/load, custom Annotations, Generics, Enums for command words, and a full Parser for typed input. Supports both GUI buttons and WASD keyboard controls, inventory management, sound effects, and pixel-art room assets.

A machine learning side project building a neural network model from scratch using NumPy. The model predicts student pass/fail outcomes based on study hours using a synthetic dataset. Implements core neural network concepts including forward propagation, loss calculation, backpropagation, and gradient descent. Built as a learning exercise to deepen understanding of Python and how neural networks function at a fundamental level. A work-in-progress project developed alongside university coursework.

A privacy-first, fully offline carbon footprint analyser built for the Snapdragon: AI Unplugged Hackathon at University of Limerick (Qualcomm × Lero, April 2026). Points your laptop camera at any everyday object and instantly displays its lifecycle CO₂ cost — sourced from a 40+ item database built on DEFRA 2023, EU JRC, and EPA Ireland data. MobileNetV3-Large handles real-time object recognition and TinyLlama-1.1B generates plain-English sustainability recommendations, both running on the Snapdragon X Elite NPU via ONNX QNN Execution Provider at ~12ms per frame. Zero network requests, zero cloud dependency — every byte of inference stays on-device.

A full-stack schema reconciliation and migration platform built for HackUPC 2026. Analyzes differences between source and target database schemas, detects conflicts such as type mismatches, missing keys, and data loss risks, and generates confidence-scored mappings. The system produces production-ready SQL migration scripts with rollback support across formats including Generic SQL, Flyway, Liquibase, and AWS DMS. Built with FastAPI backend and React + TypeScript frontend, featuring real-time collaboration, visual schema diffing, and structured migration workflows.

AI-powered exam question generator. Upload any document — PDF, slides, or notes — and instantly get AI-generated multiple choice questions. Built with a React + TypeScript frontend powered by Vite and Tailwind CSS, and a FastAPI backend that supports both Ollama (local LLMs) and the Anthropic Claude API. Features timed quiz mode, streak tracking, bookmarks, difficulty ratings, smart results analytics, and resume-interrupted quizzes via localStorage. Live with a coming soon landing page deployed on Vercel.