I am finishing my PhD at the University of Auckland under the supervision of Prof. Michael Witbrock and Prof. Gillian Dobbie. I hold a Dipl. Ing. (eq. Msc. in software engineering) at the National Institute of Applied Sciences (INSA) of Rennes and a dual Msc. in computer science and machine learning from the University of Rennes 1. Before starting my PhD, I worked as a software engineer at Alten and Amadeus IT Group as a contractor where I developed C++ software for a key component of Amadeus’s Global Distribution System, enhancing scalability of the retrieval pipeline. I also hold a research assistant position at the University of Auckland, where I lead the research and development of neural-causal networks for behaviour discovery in multi-agent systems.
Research Focus
I work on building trustworthy AI systems that can reason and understand the world around them. My main interests are on factuality, abstract and causal reasoning, which are crucial for creating robust and reliable deep learning systems in real-world production environments.
Education & Background
Research Highlights
I led the first project to evaluate and point out the brittleness and limitations of LLMs for learning robust representations of abstract concepts. I worked on showing that inducing LLMs and feedforward neural networks to behave as causal models can improve the learning of robust and domain-invariant mechanisms, building and fine-tuning a novel modular language model architecture based on causal principles for out-of-distribution abstract reasoning. I also created a novel variational auto-encoder based on latent space quantization and causal mechanisms. I built the first end-to-end framework for causal extraction and inference with large language model agents. I am currently expanding this framework to enable LLMs to retrieve causal structures from natural language and harness them for knowledge discovery and reliable reasoning. Finally, I led the research and development of an open-source project building the first interpretable neural-causal network for behaviour discovery in multi-agent systems.
Publications
My research has been featured at CORE A* AI and machine learning conferences, including IJCAI, AAMAS, EMNLP, AAAI and ACL, and high-impact workshops such as AGI@ICLR, CV4Animals@CVPR, CaLM@NeurIPS, and CRL@NeurIPS. Additionally, I serve as a reviewer for leading conferences like ICLR, EMNLP, ECML PKDD, and COLM. Find the complete list of my publications on my Google Scholar profile.
Awards & Recognition
In 2023, I was honored with the University of Auckland Best Student Published Paper in Computer Science award for my paper on disentanglement with causal interventions. In 2024, I had the opportunity to present my work at the IJCAI 2024 doctoral consortium and served as a panelist for the Global Sustainable Development Congress (GSDC) 2024 alongside Profs. Siah Hwee Ang (Chair in Business in Asia at Victoria University of Wellington), Low Teck Seng (Senior Vice President at National University of Singapore), and Banchong Mahaisavariya (President of Mahidol University). I also featured in the University of Auckland InSCight journal, which highlights the achievements of students and staff in the Faculty of Science, and delivered a talk at the Natural, Artificial and Organizational Intelligence Institute (NAOInstitute) symposium on Creativity and Intelligence. I am also a recipient of the DAAD AInet Fellowship for 2024 - AI for science.
Research Schools
In November 2020, I attended the research school on Contraint Programming, Combinatorial Optimization and Machine learning, jointly organized by the CNRS’s Laboratory of Analysis and System Architectures (LAAS-CNRS), the Association for Constraint Programming (ACP), the Operational Research and Artificial Intelligence research groups (GDR RO and GDR IA), and the Artificial and Natural Intelligence Toulouse Institute. I learned about the latest advancements in constraint programming for machine learning and finished at the 4th place in the organized hackaton. In March 2021, I attended the Reinforcement Learning Virtual School organised by the Artificial and Natural Intelligence Toulouse Institute and in September 2021, I once again attended the CNRS school on AI and Explainability, organised by the Pierre and Marie Curie University.