My long-term research goal is to gain a deeper understanding of the principles that govern life and, more generally, any class of systems that appear to self-regulate. I approach this problem by studying the interaction between an _agent_ and its environment as an integrated whole (e.g., the perception-action loop). During my PhD, I used machine learning, probabilistic, and information-theoretic tools to develop learning schemes that enable agents to build internal representations and world models to mediate their interaction with their environments, integrating these models with planning and reinforcement learning algorithms. Other interconnected research interests include:


  • The nature of information-processing: Information plays a central role in understanding living systems and agents more generally. It has structure and comes in different forms. Not all information is equivalent for an organism; some of it is semantically meaningful, and some is more valuable or relevant than other kinds. In this sense, information serves as a resource required for prediction, decision-making, and behavior. I am particularly interested in using an information-processing lens to characterize the internal organization of an agent and to interpret its dynamics in terms of information and its transformations.

  • The energetics of agent-environment coupling: Organisms operate far from thermodynamic equilibrium, continuously harvesting and consuming energy to sustain their identities. Novel developments in stochastic thermodynamics, together with their growing connections to information theory, provide new tools for understanding the energetic constraints shaping an agent’s interactions.

  • Biology of cognition: In cognitive science and the philosophy of mind and life, concepts such as agency, autonomy, adaptivity, precariousness, closure, selfhood, and autopoiesis have been developed to understand what makes a system a genuine agent. To what extent can these notions be formalized? How do they arise from underlying biological organization? Are they graded properties that apply to a wider class of systems?