Basis logotype

Markus Heinonen

People

Markus Heinonen

Research Scientist, New York, NY

Markus Heinonen is a research scientist at Basis working on Bayesian deep learning, generative modeling, and the learning of dynamical systems. His research spans diffusion models and normalizing flows, neural ODEs, SDEs, and PDEs, Gaussian processes, and the broader questions of uncertainty, calibration, and priors in deep models, with applications across the physical and life sciences.

Website Google Scholar GitHub Email

Education

  • PhD, Computer Science
    University of Helsinki, 2012

Previous Appointments

  • Research Fellow
    Aalto University
  • Postdoctoral Researcher
    Télécom ParisTech

About

I am a research scientist at Basis, where I work on probabilistic and generative machine learning for systems that evolve over time. My research centers on Bayesian deep learning, generative modeling with diffusion models and normalizing flows, and the learning of dynamical systems described by ODEs, SDEs, and PDEs.

I am especially interested in models that quantify their own uncertainty and respect the structure of the problems they are applied to—through well-chosen priors, calibration, and principled model selection. Much of my work brings these ideas to bear on scientific domains, from climate forecasting to molecular and antibody design.

Before joining Basis, I was a research fellow at Aalto University in Finland, working on generative AI, dynamical systems, and Bayesian learning. I was previously a postdoctoral researcher at Télécom ParisTech with Florence d’Alché-Buc, focusing on Gaussian processes, kernel methods, and dynamics. I completed my PhD in computer science at the University of Helsinki in 2012, advised by Juho Rousu, on kernel methods for metabolomics.

Projects

Current and recent Basis projects.

Core Technology

Foundational methods for reasoning, probabilistic inference, and model-building in complex environments.

R-ADA

A rational automated design agent for robotics, combining language models, simulation, probabilistic programming, and Bayesian inference.

Cities

Participatory city models that help residents, community groups, and policymakers reason about uncertain causes and policy consequences.

Collaborations

Ways to work with Markus at Basis.

Collaborative research projects

I am keen to collaborate with methodologists advancing generative modeling, Bayesian deep learning, and the learning of dynamical systems, as well as with scientists who want to apply these methods to problems in the physical and life sciences.

Recent Publications

Recent papers and preprints.