AI & Machine Learning Engineer building production-grade Generative AI, NLP, and agentic systems.
Currently an AI Intern at GBG, where I develop LLM applications and multi-agent workflows with a focus on scalable, real-world impact.
I bridge advanced research with practical deployment, combining an Mechatronics Engineering background with a strong interest in robotics, edge AI, and intelligent automation.
🎯 Currently focused on:
LLMs RAG Pipelines LangGraph Multi-Agent Systems Multimodal AI Production Observability
2026 ──▶ AI Trainee @ Global Brands Group (GBG)
└─ Multi-agent systems · RAG · Azure AI Stack · Langfuse · Arize Phoenix
2026 ──▶ Software Trainee @ SEITech Solutions
└─ Embedded Linux · Functional Safety · SCADE · Automotive SDV
2025 ──▶ AI Trainee @ National Telecommunication Institute (NTI — Hire-Ready)
└─ LLMs · CNNs · Vision Transformers · Fine-tuning (LoRA / PEFT)
2025 ──▶ Data Science & Gen AI @ Digital Egypt Pioneers Institute (DEPI)
└─ RAG Medical Assistant · LLMs · Semantic Search · Multilingual NLP
2020–2025 B.Sc. Mechatronics Engineering · Fayoum University
└─ GPA 3.08 · Graduation: Gloves Teleoperation Robot — A+ · ITIDA/ITAC Sponsored
| 🧩 | Agent Orchestration | Robust state machines, HITL approval, multi-agent fallback recovery |
| 📐 | RAG Evaluation | Retrieval quality metrics, confidence gating, hallucination detection frameworks |
| 🌐 | GraphRAG | Neo4j + Leiden algorithm + community-aware retrieval for knowledge graphs |
| 🏭 | Production MLOps | Observability-first pipelines with Langfuse, Arize Phoenix, and structured logging |
| 🎙️ | Multimodal Pipelines | Audio, video, and document intelligence — failure-mode-aware architecture |
"I don't stop at a working prototype.
I iterate until I understand the trade-offs, the failure modes,
and why a choice is better than its alternatives."

