PeopleofSuperintelligence

Researcher

Björn Lütjens

Björn Lütjens is a machine learning researcher who develops neural emulators and benchmarks for weather, climate, and Earth system models. After earning a B.Sc. in Engineering Science at the Technical University of Munich, he completed an S.M. in autonomous systems at MIT in 2019, where he worked on safe and robust deep reinforcement learning with Jon How and Michael Everett. He defended his MIT doctoral thesis on multiscale neural operators for climate projections with Dava Newman in May 2023 and coauthored ClimSim, which won a NeurIPS 2023 Outstanding Dataset Paper award. In August 2024, he and his MIT collaborators published a study on the ClimateBench benchmark showing that linear regression outperformed ClimaX on three of four climate variables. As of October 2026, he is a Research Scientist in Algorithms and Applications on the NASA team at IBM Research in Yorktown Heights.

Badges

Timeline

  1. 2018

    Navigation in pedestrian crowds

    At MIT's Aerospace Controls Laboratory, he studies robust and interpretable AI, safe reinforcement learning, and Bayesian deep learning. In 2018, he and Michael Everett develop a reinforcement learning framework for robust and interpretable navigation in pedestrian crowds to address overconfident neural network failures on novel observations.

  2. 2019

    Autonomous systems master's at MIT

    He earns an S.M. in Aeronautics and Astronautics in autonomous systems at MIT in 2019, working with Jon How on safe and robust deep reinforcement learning. His earlier education includes a B.Sc. in Engineering Science with Emmanuel Dean at the Technical University of Munich (TUM).

  3. 2020

    Certified adversarial robustness for deep RL

    With Michael Everett at MIT's Aerospace Controls Laboratory, he develops deep reinforcement learning algorithms in 2020 for certified adversarial robustness against perturbations in the observation space. In 2020, he also works on satellite imagery visualizing coastal floods and melting Arctic sea ice, presented at the NASA/SETI Frontier Development Lab with support from MIT Portugal, NASA, Google Cloud, and Microsoft AI for Earth.

  4. May 2023

    Doctoral thesis on climate emulators

    His doctoral thesis, Deep Learning Emulators for Accessible Climate Projections, is scheduled for defense in May 2023 as he completes a Ph.D. with Dava Newman in MIT AeroAstro on machine learning and Earth system modeling. The thesis explores how deep learning can make climate models more accessible, introduces a multiscale neural operator that reduces runtime complexity from quadratic to quasilinear along with physically consistent satellite imagery, and wins a best student presentation award at AGU in 2023.

  5. 2023

    The ClimSim climate dataset

    In 2023, he coauthors ClimSim, a multi-scale dataset of 5.7 billion pairs of multivariate input and output vectors that isolate high-resolution physics for hybrid ML-physics climate research. NeurIPS recognizes ClimSim as an Outstanding Dataset Paper in the 2023 Datasets and Benchmarks Track.

  6. 2024

    Benchmarking climate emulators

    On the ClimateBench benchmark, he, Raffaele Ferrari, Duncan Watson-Parris, and Noelle Selin find in August 2024 that linear regression outperforms the 100M-parameter ClimaX foundation model on three of four regionally resolved climate variables, including surface temperature and precipitation. Their 2024 study also raises the benchmark targets from 3 to 50 climate simulations per emissions pathway from the MPI-ESM1.2-LR model, showing that deep learning can outperform linear pattern scaling for precipitation.

  7. March 2025

    Ocean engineering seminar at MIT

    In March 2025, he gives an Ocean Engineering Seminar at the MIT Center for Ocean Engineering on the impact of internal variability on benchmarking deep learning climate emulators.

  8. As of October 2026

    Research Scientist at IBM Research

    Following Ph.D. and postdoctoral work at MIT with Dava Newman, Raff Ferrari, and Noelle Selin on machine learning for Earth system models, he is a research scientist at IBM Research by July 2025, developing machine learning methods for weather and climate emulators. In 2025, he coauthors Advances in Emulating Earth System Models with Kalyn Dorheim at an AGU 2025 workshop. As of October 2026, he is a Research Scientist in Algorithms and Applications on IBM Research's NASA team in Yorktown Heights.

Career

  1. NowIBM ResearchResearch Scientist
  2. From 2018Massachusetts Institute of TechnologyPostdoctoral Researcher

Education

  1. Massachusetts Institute of TechnologyS.M. in autonomous systems
  2. Technical University of MunichB.Sc. in Engineering Science

From the biography.

  1. 23% of people shaping AI who left a university for a company later returned to a universityIncludes IBM. We estimate that 468 of the 2,011 people shaping AI who left a university for a company (23%) later returned to a university after a median of 5 years, and 163 of 829 (20%) made the reverse round trip.
  2. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes IBM. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  3. Universities keep AI researchers for a median of 6 years, compared with 2 years at frontier labsIncludes IBM. Across 15,811 dated work stays (6,529 people), median role tenure runs from 6 years at universities to 2 years at frontier labs. We compare all five organization types, starting decades from the 1970s to the 2020s, and the organizations that keep people longest.
  4. People shaping AI work at a median of 3 organizations over a career: 3 for university starters and institute and lab starters, versus 2 for company startersIncludes IBM. Across 6,853 people shaping AI with a dated work role, we estimate a median of 3 parent organizations over a career (mean 3.6), comparing role groups, starting sectors and starting decades from the 1950s to 2020–2026.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Björn Lütjens yet. Send us a free-licensed photo.

  1. Björn Lütjens - Aerospace Controls Laboratory acl.mit.edu. Checked 21 October 2026.
  2. Björn Lütjens - Postdoc at MIT blutjens.github.io. Checked 21 October 2026.
  3. Björn Lütjens - IBM Research research.ibm.com. Checked 21 October 2026.
  4. HSL Thesis Defense: Björn Lütjens (1p, 5/3, Rm 33-116) – Home hsl.mit.edu. Checked 21 October 2026.
  5. ClimSim: A large multi-scale dataset for hybrid physics-ML climate emulation papers.neurips.cc. Checked 21 October 2026.
  6. 16 Papers Accepted To NeurIPS 2023 cs.columbia.edu. Checked 21 October 2026.
  7. The impact of internal variability on benchmarking deep learning climate emulators arxiv.org. Checked 21 October 2026.
  8. Student Seminar: The impact of internal variability on benchmarking deep learning climate emulators – Center for Ocean Engineering oe.mit.edu. Checked 21 October 2026.