A bit about me.

I'm Moroccan, and my journey started in Morocco before it took me to Germany. Today, I'm doing an M.Sc in Computer Science (Machine Learning) at TUM.

Before university, I completed a French Baccalaureate at Al Jabr School in Casablanca and worked my way up to German B2 before moving to Germany. I then completed Studienkolleg at the University of Paderborn and started my B.Sc at TUM at the end of 2021.

My published research on LLMs has been cited 170+ times. The paper provides early empirical evidence from real research workflows, especially code generation and data-analysis scripting, and shows both strong potential and clear integrity gaps across tools.

At the Fraunhofer Society, I worked on a private RAG pipeline for institutional knowledge, with a strong focus on accuracy, security, and guardrails.

At Allianz SE, I built GenAI applications for insurance operations, from customer communication analysis to agentic workflows, and I helped make those systems understandable for non-technical teams.

At BCG Platinion, I work on AI strategy in the energy space, from solution architecture and feasibility work to building things teams can actually use with clients.

My bachelor's thesis, Echo, gave me a practical foundation in hallucination mitigation, and starting my master's at TUM was the natural next step to go deeper into reliable AI systems.

I'm drawn to problems where AI needs to be trusted, not just accurate. Regulated industries — insurance, energy, healthcare — where a wrong output has real consequences. That's where interesting engineering happens.

Beyond client work, I'm fascinated by multi-model orchestration, hallucination mitigation, and building AI systems that explain their reasoning. My personal project, Quetzal, is an AI-native learning platform that combines multi-model AI with pedagogical theory. It's where I push my own limits.

I also contribute to Atlas Munich, a student initiative supporting Moroccan students in Germany. I contributed to the AI-related parts of the platform and see it as a social contribution to help new students navigate their journey with more clarity and confidence.

I don't believe in AI for AI's sake. Every system I build must answer three questions: Does it solve a real problem? Can the people who need it actually use it? Will it still work when I'm not in the room?

I work iteratively, communicate proactively, and design systems with the assumption that they need to be understood, maintained, and trusted by people who didn't build them.

When a project isn't right for AI, I'll say so. Honestly and early. That clarity is more valuable than any prototype.

I'm building toward a career at the intersection of AI research and applied engineering. Whether that's in enterprise AI, AI safety research, or building my own products, the common thread is the same: make AI work reliably for people who depend on it.