Hi, I'm Conor 👋

I'm a researcher and engineer based in Berlin, working at the intersection of probabilistic inference, deep learning, model-based reinforcement learning, and multi-agent systems. More broadly, I work on deep problems around bringing uncertainty and probabilistic principles into the study of natural and artificial intelligence.

In the past I worked at Nested Minds, a startup focused on bringing active inference to real-world applications, and at VERSES AI , where I was Technical Lead in the ML Foundations Lab — working on ways to hybridize the best of Bayesian methods (uncertainty quantification, continual learning, sample efficiency, structured prior knowledge) with the scale and expressivity of modern deep learning architectures like transformers and state space models. While in that role I co-led the development of AXIOM , which learns to play Atari-like games in minutes by growing its own object-centric structure online. I'm the lead developer and maintainer of pymdp — the de facto open-source library for active inference in discrete state spaces, with 718 GitHub stars and an active user base spanning academia and industry.

I also consult on Bayesian modeling and on hardware-aware optimization of Bayesian inference algorithms for embedded deployment.

I did a PhD at the Max Planck Institute of Animal Behavior and the University of Konstanz, advised by Iain Couzin and Karl Friston , where I modeled collective behavior through the lens of free-energy minimization. This doctoral research allowed me to synthesize my interests in collective animal behavior with the Bayesian brain, probabilistic machine learning, while also getting the chance to build highly optimized, vectorized code for large-scale multi-agent simulations and parallelize them on GPU clusters. This experience set me up nicely for my current interests in researching, building, and deploying probabilistic ML models. Before that I studied neuroscience: an M.Sc. at the University of Göttingen through the IMPRS for Neurosciences (Max Planck), and a B.A. at Swarthmore College.

Recent publications
See all publications
  • 2025 · arXiv preprint
    AXIOM: Learning to Play Games in Minutes with Expanding Object-Centric Models
    Conor Heins, Toon Van de Maele, Alexander Tschantz, Hampus Linander, Dimitrije Markovic, Tommaso Salvatori, Corrado Pezzato, Ozan Catal, Ran Wei, Magnus Koudahl, et al.
  • 2025 · Frontiers in Network Physiology, 5:1521963
    From Pixels to Planning: Scale-Free Active Inference
    Karl Friston, Conor Heins, Tim Verbelen, Lancelot Da Costa, Tommaso Salvatori, Dimitrije Markovic, Alexander Tschantz, Magnus Koudahl, Christopher Buckley, Thomas Parr
  • 2025 · arXiv preprint
    Soft Geometric Inductive Bias for Object-Centric Dynamics
    Hampus Linander, Conor Heins, Alexander Tschantz, Marco Perin, Christopher Buckley
  • 2024 · PNAS, 121(17): e2320239121
    Collective Behavior from Surprise Minimization
    Conor Heins, Beren Millidge, Lancelot Da Costa, Richard Mann, Karl Friston, Iain Couzin
Get in touch

The best way to reach me is by email, or via any of the socials below.