Francisco Girbal Eiras

Francisco Girbal Eiras is a PhD in Engineering Science (AIMS CDT) from the University of Oxford, with a focus on optimization and robust machine learning. He currently works as an ML Research Scientist at Dynamo AI since 2024, leading agentic red-team evaluations for LLMs. Prior roles include Research Scientist at FiveAI (2018–2020), where he developed motion planning algorithms and contributed to robust deep learning research.

Education highlights:

• BSc in Electrical and Computer Engineering from Técnico Lisboa (18/20, Top 2%)
• MSc in Computer Science from University of Oxford (Distinction)
• PhD in Engineering Science (AIMS) from University of Oxford (2020–2024, funded by FiveAI)

Research interests center on optimization, robust machine learning, and neural network verification. His work bridges autonomous driving and formal verification, with applications to physics-informed neural networks and LLM safety.

Key contributions include:

• Advancing zero-shot referring segmentation (state-of-the-art results at FiveAI, 2023)
• Extending randomized smoothing to anisotropic certificates (FiveAI, 2021)
• Developing correct-by-construction driver assistance systems (University of Oxford, 2017–2018)

Scientific recognition:

• Outstanding Paper Award at ICML 2023 Workshop on Formal Verification of Machine Learning
• Academic Excellence Awards at Técnico Lisboa (2013–2016)
• Swiss-European Mobility Programme Scholarship during EPFL exchange (2016–2017)

Technical expertise spans motion planning, constrained optimization, LLM evaluation pipelines, and formal verification of neural networks. He collaborates with research groups like the Oval Group and Torr Vision Group at Oxford.