AI Engineering Student · Researcher · Builder
I'm a fourth-year AI & Data Science Engineering student at the National Higher School of Artificial Intelligence (ENSIA) — one of Algeria's most selective institutions dedicated entirely to AI. My work lives at the intersection of deep research and hands-on engineering: from designing novel architectures and running training pipelines, to writing low-level GPU kernels and shipping real systems.
I care about building things that are not just accurate, but principled — systems grounded in theory, robust under pressure, and useful in practice.
Scientific Machine Learning · Physics-Informed Neural Networks · PDE Solving
Reinforcement Learning & Control · Agentic AI Systems · LLM Infrastructure
NLP for Low-Resource Languages · Privacy-Preserving AI · Efficient Deep Learning
Kernel Optimization · Homomorphic Encryption · Computer Vision
Current directions include robust RL policies for real-world control under stochastic disturbance, building an agentic AI framework running local LLMs on top of an OS, and applying agentic search strategies (à la AlphaEvolve / AdaExplore) to automate Triton kernel optimization.
PyTorch · TensorFlow · Keras · Scikit-learn · OpenCV · Gym / Stable-Baselines3
Triton · CUDA (basics) · Celery · Redis · Firebase · Jupyter · NumPy · Pandas
| Project | Description | Stack |
|---|---|---|
| Wind-Robust RL for Drone Delivery | PPO agent achieving 90–95% success under stochastic wind disturbance; OU-process wind model + curriculum training | PyTorch, PyBullet, SB3 |
| Physics-Aware Transformer for PDEs | Hybrid FNO + Transformer with physics-informed constraints for stable autoregressive PDE solving under OOD conditions | PyTorch |
| Neural Network Compression Eval System | Secure, distributed Discord bot for AI datathon: sandboxed Docker execution, async Celery + Redis pipeline | Docker, Redis, Flask |
"The best way to understand something deeply is to try to break it, rebuild it, and explain it to someone else."