Djalel Benbouzid

Machine Learning & Trustworthy AI

I build Trustworthy AI systems by developing robust governance, ensuring EU AI Act compliance, and creating production-ready generative models.

Professional Profile

With over 15 years of experience as an AI practitioner, my work is rooted in deep technical research, from foundational reinforcement learning to the development of generative models. At Volkswagen Group, I applied this background to build the company's first enterprise-wide AI governance system and implement full compliance with the EU AI Act a year ahead of schedule.

Today, I remain deeply engaged with the technical frontier, focusing on LLM architectures and multi-agent systems. To ground this work in real-world outcomes, I am an active contributor to the AI Incident Database (AIAAIC), where I analyze system failures to help inform safer development. This work is central to my approach: evaluating emerging technologies through a dual lens of technical feasibility and ethical responsibility.

Beyond my corporate and research roles, I am passionate about fostering the next generation of talent. This commitment drives my work as a data science teacher and my active involvement in organizing and lecturing at machine learning summer schools across the Global South.

Areas of Expertise

Reinforcement Learning (15 years)

  • PhD thesis with applications to high-energy physics
  • Multiple publications (ISRR, IROS, ICML)
  • Led specialized RL research team at VW Group
  • Lectured at ML summer schools across 2018-2019

Generative AI (7 years)

  • Developed generative models for theatrical production
  • Published novel approaches to creative AI
  • Developed POCs for agentic-LLM applications

AI Ethics & Governance (6 years)

  • Co-founded & managed ETAMI EU consortium
  • Published in FAccT and AIES
  • Developed and implemented the company's first compliance methodologies for the EU AI Act
  • Active contributor to the AI Incidents Database

Selected Publications

Speaking

Keynote on the nuanced risks of current LLM architectures and the potential of agentic systems, delivered to 3000+ industry leaders at Generation AI.

Generation AI, Paris - Dec 2024

Keynote on Generative AI, AI Act compliance, and novel solutions

Generative AI EU, Amsterdam - Dec 2024

Additional Experience & Service

Education

PhD in Machine Learning

University of Paris-Saclay

2010 - 2014