██████╗ ██████╗ ██╗ ██╗ ██████╗ ██╗ ██╗██╗
██╔══██╗██╔═══██╗██║ ██║██╔════╝ ██║ ██║██║
██████╔╝██║ ██║██║ ██║██║ ███╗██║ ██║██║
██╔══██╗██║ ██║██║ ██║██║ ██║██║ ██║██║
██║ ██║╚██████╔╝╚██████╔╝╚██████╔╝╚██████╔╝██║
╚═╝ ╚═╝ ╚═════╝ ╚═════╝ ╚═════╝ ╚═════╝╚═╝
Systems that reason, act, and learn — from bare metal CUDA kernels to full-stack agentic platforms deployed in production.
I'm an AI Engineer and Researcher, specializing in the full vertical of modern AI:
- Agentic Frameworks — multi-step reasoning pipelines that diagnose, act, and recover autonomously
- Deep Learning — CV, NLP, Reinforcement Learning, and custom model architectures
- LLMs in Production — RAG systems, observability, evaluation, and fine-tuning
Currently building AI at Nemo Technologies and taking on freelance AI projects globally.
LLM-powered framework for automated diagnosis of data pipeline failures
- Fuses data lineage, logs, metrics, and git insights via LangChain + Qdrant + OpenLineage
- Incident MTTD reduced from hours → 22 seconds
- Full-stack: Next.js frontend · FastAPI microservice · PostgreSQL · Langfuse observability
Python Next.js LangChain Qdrant FastAPI PostgreSQL Langfuse
Real-time medical knowledge delivery for pediatric doctors
- Production-ready pipeline: semantic + vector search, hallucination reduction
- Evaluation via RAGAS; monitoring via Langfuse
- Stack: Next.js · Vercel AI SDK · Gemini · Redis · Docker
LangChain Gemini Next.js PostgreSQL Redis RAGAS Docker
Math tutoring SaaS with personalized learning paths and quiz generation
- Semantic search with Qdrant + Gemini embeddings
- Custom workflow engine for adaptive quizzes and exam prep
- Deployed on Vercel · Backend: FastAPI
Next.js FastAPI Qdrant Gemini Vercel AI SDK
3D skeleton-based action recognition · 94% validation accuracy
- Shifted from ST-GCN → Hybrid CNN-LSTM for temporal pattern recognition
- Built with PyTorch + MONAI · Tracked via WandB
- TDD, Clean Code, reproducible MLOps pipeline
PyTorch MONAI WandB scikit-learn Git
AI/ML PyTorch · TensorFlow · scikit-learn · LangChain · Hugging Face
Agno · Vercel AI SDK · MONAI · Gemini · OpenRouter
Engineering Python · TypeScript · Next.js · FastAPI · SQL
PostgreSQL · MongoDB · Streamlit · Redis
MLOps Docker · WandB · DVC · Hydra · Prometheus · Grafana
Langfuse · RAGAS · UV · Conda · CML
Research Multi-Agent RL · RLHF · CUDA Acceleration · Autonomous Navigation
| Area | Focus |
|---|---|
| Autonomous Systems | Multi-Agent RL for USV path planning & autonomous navigation |
| LLM Alignment | RLHF and its transposition to autonomous maritime systems |
| DL Acceleration | CUDA/PyTorch optimization — achieved 200× speedup on embedding models |
- 🥇 Smart Trading Challenge — Winner; hybrid AI system for ship tracking & situational awareness
- 🌍 MedGemma Global Challenge — Offline, edge-oriented agentic medical solution
- 📜 Hugging Face — AI Agents & MCP certified
- ✍️ Blog — Accelerating Deep Learning with PyTorch and CUDA · Read →
- 🎓 Instructor — ML, CUDA for AI Engineers, LLMs, Mathematics for ML (2024–2026)
- 🔭 Research — French Naval Academy (maritime surveillance AI) · COGE (5G comms + RL driving)
Arabic Native · French Native · English Professional
Building AI that acts, not just predicts.