José Pombal is a Senior AI Research Scientist at Sword Health, where he focuses on large language models and the evaluation of AI systems for healthcare. His research addresses one of the most consequential questions in clinical AI. Not just whether a model performs well on a benchmark, but whether it can be trusted to perform consistently and safely in the environments where it actually matters.
At Sword, José contributes to research that ensures language models are evaluated with the rigour required before they support clinical workflows and member interactions. His work strengthens the scientific foundation behind Sword’s approach to generative AI, building the measurement and evaluation capability that allows the platform to grow without compromising on quality or safety.
José brings experience from both academic research and applied industry settings. Before joining Sword, he worked as an AI Research Scientist at Unbabel, where he focused on language model research, and as a Machine Learning Research Intern at Feedzai, where his work centered on responsible AI, including published research on fairness and bias in high-stakes machine learning applications. He approaches AI evaluation with the discipline of someone who understands that in healthcare, the cost of an unreliable system is not an engineering problem. It is a care problem.
Education
José holds a Master’s degree in Data Science and Engineering from Instituto Superior Técnico in Lisbon and a Bachelor’s degree in Economics from Nova SB. His master’s thesis on algorithmic fairness in machine learning received a grade of 20/20. He is currently a Computer Science PhD candidate at Instituto Superior Técnico.
Experience
Before joining Sword Health, José was an AI Research Scientist at Unbabel, where he worked on language model research for over two years and published several works at top-tier venues like TACL, COLM, EMNLP, and ACL. Prior to that, he completed a machine learning research internship at Feedzai focused on responsible AI, with published work at NeurIPS examining fairness and bias in machine learning systems. At Sword, he focuses on frontier AI research, with particular emphasis on LLM evaluation and the development of trustworthy generative AI systems for healthcare.
