Jonas Kubilius

Timeline

1. 2005–2012

   Visual cortex at MIT and Leuven

   He earns a BSc in mathematics and physics at MIT and an MSc and PhD in artificial intelligence and psychology at KU Leuven. From 2005 to 2012, he studies how the human visual cortex represents shapes, objects and scenes, with support from an FWO fellowship.

2. 2014–2016

   Shape in deep neural networks

   In December 2014, he publishes a conceptual framework of computations in mid-level vision with Johan Wagemans and Hans Op de Beeck at KU Leuven's Laboratory of Biological Psychology. In 2016, with Stefania Bracci and Op de Beeck, he shows that deep neural networks trained for generic object recognition develop sensitivity to shape features characteristic of human and primate vision.

3. 2017–2018

   A Marie Curie fellowship at MIT

   By November 2017, he is a postdoctoral researcher at MIT and KU Leuven on a Marie Skłodowska-Curie fellowship, developing deep learning models of the primate visual system at MIT. He coauthors NeurIPS papers on goal-driven models of the rodent whisker-trigeminal system in 2017 and on recurrent convolutional models of the visual system in 2018.

4. 2017–2022

   Swamp Intelligence installations

   With Nomeda and Gediminas Urbonas, he develops Swamp Intelligence, a series of AI installations made from 2017 to 2022. Its sngan_projection network, which generates hybrid images from white noise, is part of the Swamp Pavilion in 2018 and appears at the Taipei Biennial in 2020.

5. 2019

   Recurrence in the ventral stream

   In 2019, he coauthors a Nature Neuroscience paper with Kohitij Kar, Kailyn Schmidt, Elias Issa and James DiCarlo. Using behavioral methods and large-scale electrophysiology, the team argues that recurrent circuits in the primate ventral stream are critical for rapid object identification.

6. 2019

   Brain-Score and CORnet-S

   At MIT's Center for Brains, Minds and Machines, he and Martin Schrimpf lead work on Brain-Score, a way to check how well models match the brain, and on CORnet-S, a shallow network with four anatomically mapped areas and recurrent connectivity. Their paper on CORnet-S is presented as an oral at NeurIPS 2019.

7. 2019–2024

   Co-founding Three Thirds

   From 2019 to 2024, he is co-founder and CEO of Three Thirds, an AI solutions firm working on real-time object detection, medical image segmentation, production ML pipelines and deep reinforcement learning. From 2022 to 2023, he is co-founder and CTO of HERlab, a startup developing novel protein expression systems through high-throughput screening of unconventional yeasts.

8. Since 2024

   Genome editing in Vilnius

   Since 2024, he engineers CRISPR-Cas systems and genome editing tools for bacteriophages at the EMBL Partnership Institute of Vilnius University's Life Sciences Center, where he is scientist in residence in Stephen Jones's lab. In 2026, he recruits a PhD student for a project that pairs experimental evolution of CRISPR-Cas enzymes with machine learning.