Real-Time Taxi Stream Processing

A distributed real-time analytics system for taxi movement data. It uses Kafka for ingestion, Storm for stream processing, and Redis for low-latency state and alerting. Results are streamed to a live dashboard over WebSockets.

Tech: Kafka · Storm · Redis · Docker · WebSockets

Rust OS – Minimal x86_64 Operating System

A minimal operating system written in Rust for x86_64. It boots in QEMU, sets up paging, interrupts, and memory management, mounts a TAR-backed RAM disk at /, and provides an interactive shell with tab completion and file browsing.

The shell can execute WASM applications that render directly to the framebuffer and return control to the OS, demonstrating user-kernel interaction and system extensibility.

Tech: Rust · x86_64 · QEMU · Paging · Interrupts · WASM · Framebuffer

TutorMatch

TutorMatch is a platform I built with university colleagues to connect students for collaborative learning. It helps students find study partners, organize sessions, and learn from each other more effectively.

Shazam - Audio Retrieval

This project implements the audio identification approach from the original Shazam paper (Wang, 2003). It uses audio fingerprinting and hash-based matching for fast and robust song retrieval. By extracting distinctive spectral features, the system builds compact fingerprints that can identify songs from short, noisy, or distorted recordings.

Rocket League - Trick Shot Classification

This project was part of a university Kaggle competition. I developed a machine learning model to classify Rocket League trick shots based on in-game metrics and player inputs. Using a dataset of 297 trick shots, I combined time-series features and summary statistics to train and evaluate a classifier for different maneuver types.

Distributed Message Broker Service

A distributed message broker service built for reliable and scalable communication. Brokers register through a custom DNS for dynamic cluster management and route messages through exchanges and queues. Leader election with Raft, Ring, or Bully algorithms keeps the system available during failures.

Volume Visualization

This application processes medical data, including MRI and CT scans, using direct volume rendering techniques. It uses Three.js and GLSL shaders for detailed 3D visualization, while D3.js presents related analytical data to support exploration and interpretation.

Dependency Analysis

This tool tracks dependency updates across Java (Maven and Gradle), Python, and JavaScript manifest files. It analyzes how quickly updates are introduced and propagated through a codebase, and reports patterns in dependency behavior over time.