CS Student · AI & Backend Developer · Game Dev · UI/UX Designer
Building things that actually work — from ML models to pixel art games
I'm a Computer Science student at ESI-SBA with a broad builder's mindset. I work across the full stack — from training ML models and building APIs, to designing UIs in Figma and making games in Love2D. I participate actively in hackathons and love turning ideas into shipped products.
- Focused on AI/ML, backend development, and MLOps
- Indie game developer using Lua + Love2D with pixel art in Aseprite
- UI/UX designer with Figma
- Based in Algeria
AI & Data
Backend
Frontend
Game Dev & Design
1st place at MobAI'26 hackathon. Full AI system forecasting demand for 1,129 SKUs and optimizing warehouse storage + picking routes. Built with XGBoost, Prophet, and FastAPI. Deployed on Azure with 12 API endpoints.
Python XGBoost Prophet FastAPI Azure MLOps
Kaggle competition entry achieving Val Dice 0.9814 and Val IoU 0.9641 on a boundary-heavy segmentation task. The competition scores 80% on boundary precision — making standard segmentation approaches insufficient. Built a custom boundary-aware loss (BCE + Dice + Gradient + Sobel), trained a U-Net with EfficientNet-B3 encoder at 512×640px, and applied TTA + threshold optimization using the actual competition metric. Converged in ~1 hour on a T4 GPU.
PyTorch U-Net EfficientNet Computer Vision Kaggle
MLOps project using Isolation Forest to detect at-risk students. Includes experiment tracking with W&B, model versioning, CI/CD with GitHub Actions, and a Flask REST API for real-time predictions.
Python scikit-learn Flask W&B GitHub Actions MLOps
A full-stack productivity web app with a 7-day calendar view, color-coded commitments, priority tracking, and Supabase backend. Also packaged as a desktop app with Electron. Live at v0-life-style.vercel.app.
React Tailwind Supabase Vite Electron Vercel
🥈Binary X-ray classification using two independent deep learning strategies, both trained on the same dataset and submitted to the Kaggle leaderboard this one ranked second in the datathoon .
"Ship it, learn from it, improve it."