Artificial intelligence researcher in Luxembourg
I conduct artificial intelligence research at the Luxembourg Institute of Science and Technology (LIST), one of Luxembourg's leading technology research organizations.
My work focuses on natural language processing (NLP), large language models (LLM), retrieval-augmented generation (RAG), automated fact-checking and model reliability evaluation.
In parallel, I contribute to Devana.ai, a generative AI platform specialized in source verification, deployed among others at the Banque Internationale à Luxembourg (BIL).
Publications
LLM-FACETS: a privacy-preserving open-source framework for evaluating LLM transparency and accountability — token-level log-probability visualization, multi-judge consensus, RAG Triad, BLEU, ROUGE and BERTScore metrics, aligned with the EU AI Act and the NIST framework.
Read on arXiv (2605.31167) →Devana × LIST × University of Luxembourg partnership
Since 2025, Devana, LIST and the University of Luxembourg have been bound by a scientific partnership on explainable AI systems for critical and regulated environments. I am part of the research team involved in this program, alongside German Castignani and Alessio Buscemi.
The work covers explainable AI for generative models in a RAG context (mechanistic interpretability, source attribution, reference resolution), the RAG Score™ framework for measuring the reliability of retrieval-augmented generation systems, and contributions to European AI governance standards aligned with the EU AI Act.
This research builds on LIST's AI Sandbox and its technologies: FAIRGAME, LLM Factory, MLA-BiTe and KG-RAG.
Discover the Devana × LIST partnership →Applied research & industry
My approach bridges research and the field: the methods I study at LIST (LLM evaluation, RAG, hallucination detection) directly feed my AI consulting work for companies in France and Luxembourg.
Also a freelance web and application developer, I build prototypes and full products around these technologies.
Frequently asked questions
What does Tom Lucas research at LIST?
Tom Lucas researches artificial intelligence at the Luxembourg Institute of Science and Technology (LIST): natural language processing (NLP), large language models (LLM), retrieval-augmented generation (RAG), automated fact-checking, model reliability evaluation and explainable AI for generative models in a RAG context.
What is LLM-FACETS?
LLM-FACETS is a privacy-preserving open-source framework for evaluating the transparency and accountability of large language models. Published on arXiv (2605.31167) in May 2026 by Tom Lucas, Alessio Buscemi, Alfredo Capozucca, German Castignani and Barbara Delacroix, it combines token-level log-probability visualization, multi-judge consensus and RAG Triad, BLEU, ROUGE and BERTScore metrics, aligned with the EU AI Act and the NIST framework.
How is Devana.ai connected to LIST?
Since 2025, Devana, LIST and the University of Luxembourg have run a scientific partnership on explainable AI systems for critical and regulated environments. Tom Lucas is part of the research team in this program, alongside German Castignani and Alessio Buscemi.
Do you offer AI consulting to companies?
Yes. The methods studied at LIST (LLM evaluation, RAG, hallucination detection) feed directly into AI consulting work for companies in France and Luxembourg, from prototypes to full products.