Research Bio
I completed my PhD in Delft under the supervision of Dr Manuel Mazo in 2023. During these years I spent the summer of 2022 at the University of Oxford, working with Prof. Alessandro Abate on robust reinforcement learning problems. My research has ranged from mean‑field approximations for biologically inspired agents (paper) to communication‑induced uncertainty in multi‑agent systems (paper) and verifiable robustness in RL (paper). After that, I was a PostDoc Researcher at the University of Oxford, under the supervision of Michael Wooldridge, Ani Calinescu and Doyne Farmer, where I worked on risk awareness and information theoretic decision making in AI agents, and a senior research scientist at Phaidra working on world models, RL and objective mismatch.
News
- - Our proposal (co-led with Nick Bishop, David Hyland, Joel Dyer and Michael Wooldridge) has been awarded a UK AISI Alignment Fund grant! We will be working on how to develop bounded intelligence that composes into safe and general multi-agent AI systems.
- - Our paper (led by Alvaro Serra) on better priors for model-based RL has been accepted at ICML 2026! Find the preprint here.
- - Our paper on emerging risk awareness and alignment shifts for agents under resource pressure has been accepted at NeurIPS 2025! Find the preprint here.
- - We got two papers accepted at the Reliable ML from Unreliable Data workshop in NeurIPS 2025; Bayesian Decision Making Observing an Expert and Sandbagging in a Simple Survival Bandit Problem (led by Joel Dyer).
- - Our paper (co-led with Joel Dyer and Nick Bishop) on learning uninformed priors for simulation models has been accepted at ICML 2025!
- - Our paper (co-led with Giannis Delimpaltadakis) on predictability in RL agents has been accepted at TMLR!
- - Our paper (led by Roman Chiva Gil) on implicit cooperation of multi-agent planning agents via predictability awareness has been accepted at AAMAS 2025!
- - Our paper (co-led with Joel Dyer and Nick Bishop) on model exploration via entropy maximisation of marginal likelihood functions has been accepted at the NeurIPS 2024 Workshop on Data-driven and Differentiable Simulations, Surrogates, and Solvers!
- — I joined the University of Oxford as a postdoc researcher, working with Prof. Michael Wooldridge, Ani Calinescu and Doyne Farmer on ML for ABMs.
- — Our paper on lexicographic robustness in RL was selected for oral presentation at L4DC 2024.
- — I started a PostDoc position at the Cognitive Robotics department in TU Delft, working in the Autonomous Robots Lab with Dr. Javier Alonso Mora and Dr. Jens Kober on risk‑aware decision making for autonomous navigation problems.
- — Finished my PhD Thesis! You can find it here.
Links & Contact
Footnotes
1 Spanish, Catalan, English, French, Dutch
Outdoor Stuff
