Aref: an open source contributor

whoami

Aref, a full-stack developer who builds from low-level systems to polished frontends. Comfortable with a bunch of languages, always picking up new tools.

Who am I
Projects

projects

LipSync
Real-time lip reading pipeline for AR glasses. Runs a custom ONNX model on Jetson Orin, streaming decoded speech to Vuzix Z-4K.
click to expand
Python ONNX OpenCV Jetson AR
Built an end-to-end visual speech recognition system optimized for edge deployment. The pipeline captures video from AR glasses, processes frames through a lightweight lip-reading model, and streams decoded text in real-time. Trained on multiple languages with data augmentation for robustness in noisy environments.
Gilgamesh
Cross-platform hand tracking that turns your fingers into animated swords. Uses MediaPipe for hand detection with OpenCV and PyQt6.
click to expand
Python MediaPipe OpenCV PyQt6 Computer Vision
A gesture-controlled application that maps hand landmarks to animated sword trails in real-time. Leverages MediaPipe's hand tracking model with custom gesture recognition to trigger effects. Built with a PyQt6 interface for cross-platform support and OpenCV for frame processing and rendering.
Tracer
CLI tool that scans code repositories and generates Mermaid architecture diagrams. Supports Python, JS/TS, Rust, Go, Java, and C++ — no LLM required.
click to expand
Python Mermaid CLI Static Analysis
Tracer parses import statements, class hierarchies, and function calls across multiple languages to build accurate dependency graphs. It outputs clean Mermaid diagram files that can be rendered anywhere. Works with both local repositories and GitHub URLs, with optional LLM backends for enhanced summaries.
CICD Is Overrated
FastAPI app with a full CI/CD pipeline. Lint, test, coverage, type checking, security scanning, and Docker deployment to GitHub Container Registry.
click to expand
Python FastAPI Docker GitHub Actions CI/CD
A production-ready FastAPI application demonstrating enterprise-grade CI/CD practices. Features six parallel GitHub Actions jobs (lint, test, coverage, type-check, security, audit), semantic-release auto-versioning, and multi-stage Docker builds pushed to GHCR. Coverage gate set at 90% with Bandit security scanning and pip-audit vulnerability checks.

stack

Languages Go, Rust, C, Python, Java
Environment Linux (Nobara / Niri)
Shell Fish, Bash, Zsh
Editor Neovim + custom config
currently rebuilding HTTP from TCP
Tech Stack

contact

Open to opportunities, questions, or just chatting about systems — reach out!

Contact

Project Details