Jonas Adler

Timeline

1. 2017–2018

   Learned primal-dual reconstruction

   Holding an MSc in engineering physics from KTH, he researches cancer imaging at Elekta while pursuing an industrial PhD in applied mathematics at KTH. In July 2017, he and Ozan Öktem propose Learned Primal-Dual for tomographic reconstruction; the method unrolls a primal-dual optimization method and replaces its proximal operators with convolutional neural networks. IEEE Transactions on Medical Imaging publishes the paper in 2018.

2. October 2019

   A doctorate at KTH

   In October 2019, he defends “Data-driven Methods in Inverse Problems” at KTH under supervisor Ozan Öktem, with Thomas Pock of Graz University of Technology as opponent. The thesis contains six papers; its first two introduce learned iterative reconstruction methods that combine model-driven reconstruction with deep neural networks.

3. 2021

   AlphaFold in Nature

   In July 2021, he coauthors the Nature paper “Highly accurate protein structure prediction with AlphaFold,” which validates a redesigned model at CASP14. In November 2021, he explains in a KTH talk that assessors judge AlphaFold’s predictions “competitive with experiment” for about two thirds of proteins.

4. March 2024

   Gemini 1.5 technical report

   In March 2024, he coauthors the Gemini 1.5 technical report, which describes compute-efficient multimodal models that recall and reason over information across millions of tokens, including long documents, video and audio.

5. May–June 2024

   The AlphaFold 3 model

   In May 2024, he is one of eight authors credited with equal contributions to the Nature paper “Accurate structure prediction of biomolecular interactions with AlphaFold 3.” The diffusion-based model predicts the joint structure of complexes containing proteins, nucleic acids, small molecules, ions and modified residues; in June 2024, he presents it at KTH’s Digital Futures Hub.

6. October–December 2024

   A role on the Gemini team

   In October 2024, he is a Staff Research Scientist on Google DeepMind’s Gemini team, building models for language and media generation. His earlier DeepMind work covers AI for Science, and he is an AlphaFold team member. In December 2024, as a Senior Staff Research Scientist, he gives a KTH talk about his path from mathematics to AlphaFold.

7. 2025

   The Gemini 2.5 report

   In July 2025, he coauthors the Gemini 2.5 technical report, which introduces Gemini 2.5 Pro and Flash alongside Gemini 2.0 Flash and Flash-Lite. The report describes 2.5 Pro as a thinking model with state-of-the-art results on frontier coding and reasoning benchmarks.

8. As of October 2026

   Principal scientist at Google DeepMind

   In June 2026, he and Alexander Pritzel plan to leave Google for Anthropic after key roles on Gemini; Adler works on Google’s AI coding product. Both collaborate with John Jumper on his protein-folding work and plan to join him at Anthropic. As of October 2026, he is a Principal Research Scientist at Google DeepMind, working on Gemini models and machine learning for science.