Andreas Loukas
Andreas Loukas is a computer scientist whose work connects graph neural networks with AI for protein design. He earned computer science degrees at the University of Patras and TU Delft, then pursued postdoctoral research at TU Berlin and EPFL and became an assistant professor at the University of Luxembourg. At Genentech's Prescient Design, he co-led machine learning for computational molecule design, and his coauthored antibody-generation paper received an ICLR Outstanding Paper award in 2024. As of October 2026, he has a current affiliation with Isomorphic Labs and is a Principal Researcher at Google.
- Now at
- Isomorphic Labs
- Badges
- Links
- Google Scholar and LinkedIn
Timeline
-
2010–2015
Computer science degrees
He earns a computer science diploma at the University of Patras in 2010 and a computer science PhD at TU Delft in 2015.
-
2015
Distributed graph filtering
In 2015, he coauthors Distributed autoregressive moving average graph filters, a paper that introduces autoregressive moving average filters on graphs and presents a distributed implementation.
-
2016–2018
Postdoctoral research at EPFL
After postdoctoral studies at TU Berlin and EPFL, he works as a postdoctoral researcher with EPFL's LTS2 from 2016 to 2017. In 2018, he becomes an SNSF Ambizione fellow at EPFL.
-
2020
Limits of message-passing networks
In 2020, his paper What graph neural networks cannot learn: depth vs width studies the expressive power of graph neural networks in the message-passing framework.
-
2021
Assistant professor in Luxembourg
In 2021, he becomes an assistant professor in the Computer Science department at the University of Luxembourg.
-
2022–2024
Machine learning at Prescient
He joins Genentech/Roche in 2022 and co-leads the machine-learning effort at Prescient Design, developing structure-, geometry-, and constraint-aware neural networks for computational molecule design. His Roche role continues through 2024.
-
2024
An ICLR outstanding paper award
He coauthors Protein Discovery with Discrete Walk-Jump Sampling, which evaluates discrete walk-jump sampling for generative modeling of antibody proteins. The paper wins an ICLR 2024 Outstanding Paper award.
-
As of October 2026
Google and Isomorphic Labs
He is a Principal Researcher at Google, a role he holds since 2025, and has a current affiliation with Isomorphic Labs. His research spans graph neural networks and machine learning for structural and biological data.
Career
- Since 2025GooglePrincipal Researcher
- NowIsomorphic Labs
- 2022–2024RocheMachine-learning Co-Lead
- From 2021University of LuxembourgAssistant Professor
- 2016–2017EPFLPostdoctoral Researcher, SNSF Ambizione Fellow
- Technische Universität BerlinPostdoctoral Researcher
Education
- 2015Delft University of TechnologyPhD
- 2010University of PatrasComputer Science diploma
Sources and credits
Photo: We have no published photo of Andreas Loukas yet. Send us a free-licensed photo.
- Genentech: Andreas Loukas | Senior Principal Scientist and Machine Learning Lead, Prescient Design, Computational Sciences
- Distributed autoregressive moving average graph filters - TU Delft Research Portal
- Former Members ‒ LTS4 ‐ EPFL
- ICLR: What graph neural networks cannot learn: depth vs width
- Andreas Loukas (Google) — Lacuna
- Protein Discovery with Discrete Walk-Jump Sampling
- ICLR 2024 Outstanding Paper Awards – ICLR Blog





