cv
A PDF version of the CV can be downloaded by clicking the icon on the right.
Basics
| Name | Jonathan Geuter |
| [firstname].[lastname] at gmx.de | |
| Links | Google Scholar, GitHub, LinkedIn |
| Summary | PhD student in Applied Mathematics at Harvard. I work on efficient methods for training, post-training, and inference of language models across both autoregressive LLMs and diffusion LLMs. Passionate about alignment, knowledge distillation, novel architectures, diffusion models, and optimal transport. Kempner Graduate Fellow at the Kempner Institute for the Study of Natural & Artificial Intelligence. |
Education
- Sep 2023 – May 2028
Cambridge, US
PhD
Harvard University
Applied Mathematics
- Sep 2023 – May 2025
Cambridge, US
Master of Science
Harvard University
Computer Science
- En route to the PhD
- Oct 2020 – Oct 2022
Berlin, Germany
Master of Science
Technische Universität Berlin
Mathematics
- Aug 2018 – May 2019
Berkeley, US
UC Education Abroad Program
University of California, Berkeley
- Oct 2016 – Sep 2020
Berlin, Germany
Bachelor of Science
Technische Universität Berlin
Mathematics
Work
- Nov 2026 – Jan 2027
Sydney, Australia
Quantitative Research Intern (incoming)
IMC Trading
- May 2026 – Sep 2026
Sunnyvale, US
AI Research Intern
Institute of Foundation Models
- Jan 2025 – May 2025
Cambridge, US
TA for Computational Science and Engineering Capstone Project Class
Harvard University
- May 2023 – Aug 2023
Berlin, Germany
Machine Learning Intern
Jina AI
- Trained state-of-the-art open source language embedding models [Hugging Face]
- May 2021 – Nov 2022
Berlin, Germany
Research Assistant
Zuse Institute Berlin
- Worked on a Julia package for the Frank-Wolfe algorithm and GNNs in the Laboratory for Interactive Optimization and Learning
- Aug 2020 – Oct 2020
Berlin, Germany
TA for Calculus I and Linear Algebra I
Technische Universität Berlin
- Oct 2019 – Mar 2021
Berlin, Germany
TA for Computer-Oriented Mathematics I and II
Technische Universität Berlin
- Oct 2018 – Dec 2018
Berkeley, US
Research Intern
University of California, Berkeley
- Oct 2017 – Aug 2018
Berlin, Germany
TA for Computer-Oriented Mathematics I and II
Technische Universität Berlin
Publications
- Unlocking Lossless Speedups in LLMs via Discrete Diffusion, Subham Sekhar Sahoo*, Lingjie Chen*, Khiem Pham*, Jonathan Geuter*, Chaitanya Dwivedi, Varad Pimpalkhute, Yash Akhauri, Alexander Moreno, Mikhail Yurochkin, Zhenting Wang, Mostafa Elhoushi, Nolan Dey, Shane Bergsma, Joel Hestness, John Thickstun, Eric Xing, Zhengzhong Liu, arXiv preprint
- Thinking at the Right Size: Amortized Distillation Across Post-Trained LLMs, Yan Zhou, Sara Kangaslahti*, Jonathan Geuter*, Nihal V. Nayak, Marco Fumero, Francesco Locatello, David Alvarez-Melis, Findings of EMNLP 2026, Budapest, Hungary
- Understanding Layer Patching in Model Size Interpolation, Sara Kangaslahti, Jonathan Geuter*, Nihal V. Nayak*, Marco Fumero*, Francesco Locatello, David Alvarez-Melis, NeurIPS 2026, Sydney, Australia. Previously: AdaptFM Workshop at ICML 2026
- Stop Training for the Worst: Progressive Unmasking Accelerates Masked Diffusion Training, Jaeyeon Kim*, Jonathan Geuter*, David Alvarez-Melis, Sham Kakade, Sitan Chen, ICML 2026, Seoul, South Korea. Previously: MM Intelligence Workshop at ICLR 2026
- Boomerang Distillation Enables Zero-Shot Model Size Interpolation, Sara Kangaslahti, Nihal V. Nayak*, Jonathan Geuter*, Marco Fumero, Francesco Locatello, David Alvarez-Melis, ICLR 2026, Rio de Janeiro, Brazil. Previously: 3rd UniReps Workshop at NeurIPS 2025
- RoBoN: Routed Online Best-of-n for Test-Time Scaling with Multiple LLMs, Jonathan Geuter, Gregor Kornhardt, 1st FoRLM Workshop at NeurIPS 2025, San Diego, USA
- Guided Speculative Inference for Efficient Test-Time Alignment of LLMs, Jonathan Geuter, Youssef Mroueh, David Alvarez-Melis, ICLR 2026, Rio de Janeiro, Brazil. Previously: 3rd ES-FoMo Workshop at ICML 2025 (Spotlight)
- Entropy-Driven Pre-Tokenization for Byte-Pair Encoding, Yifan Hu, Frank Liang, Dachuan Zhao, Jonathan Geuter, Varshini Reddy, Craig W. Schmidt, Chris Tanner, Tokenization Workshop (TokShop) at ICML 2025, Vancouver, Canada
- Universal Neural Optimal Transport, Jonathan Geuter, Gregor Kornhardt, Ingimar Tomasson, Vaios Laschos, ICML 2025, Vancouver, Canada
- DDEQs: Distributional Deep Equilibrium Models through Wasserstein Gradient Flows, Jonathan Geuter, Clément Bonet, Anna Korba, David Alvarez-Melis, AISTATS 2025, Phuket, Thailand
- Jina Embeddings: A Novel Set of High-Performance Sentence Embedding Models, Michael Günther, Louis Milliken, Jonathan Geuter, Georgios Mastrapas, Bo Wang, Han Xiao, 3rd NLP-OSS Workshop at EMNLP 2023, Singapore
Projects
- May 2026 – Sep 2026
Institute of Foundation Models
K2 Horizon
Contributed to K2 Horizon, a suite of open-source language models ranging from 0.9B to 375B, the largest fully open-sourced LLM release of this scale
- May 2023 – Aug 2023
Jina AI
finetuner (1.5k ★)
Contributed to Jina AI's finetuner, a repository for end-to-end finetuning of embedding models
- May 2021 – Nov 2022
Zuse Institute Berlin
FrankWolfe.jl (>100 ★), CINDy
Contributed to a Julia package of the Frank-Wolfe algorithm, and a Python implementation of the CINDy algorithm
- May 2022 – Oct 2022
TU Berlin and Weierstrass Institute
- Jun 2020 – Sep 2020
TU Berlin and HU Berlin
Talks
- Aug 2025
Invited talk at Munich NLP about guided speculative inference (virtual)
- Jul 2025
Oral presentation at ICML 2025 Efficient Systems for Foundation Models Workshop about guided speculative inference
Awards
- Jun 2026
QRT ICML Travel Grant
- Apr 2026
G Research Spring into Quant Finance
selective quant finance program
- Nov 2025
IMC NeurIPS Travel Scholarship
- Jul 2025
QuantCo ICML Travel Scholarship
- Jul 2025
Hudson River Trading ICML Scholar Award
- Sep 2025
Kempner Institute Graduate Fellowship
Kempner Institute for the Study of Natural & Artificial Intelligence
- Jan 2021
Deutschlandstipendium Scholarship
Awarded for two semesters
- Jan 2020
Deutschlandstipendium Scholarship
Awarded for two semesters
- Jan 2018
Study Abroad Grant
University of California, Berkeley
Grant for two semesters of full-time study
Service
| Reviewer | NeurIPS (2023, 2025), AISTATS (2024), ICML (2025, 2026), ICLR (2025, 2026) |
Skills
| Programming | |
| Python | |
| PyTorch | |
| transformers | |
| SGLang | |
| vLLM | |
| nano-vLLM | |
| NumPy | |
| TensorFlow | |
| JAX | |
| scikit-learn | |
| pandas | |
| Lightning | |
| multiprocessing | |
| verl | |
| distributed training | |
| Julia |
| Miscellaneous | |
| Bash | |
| LaTeX | |
| Git | |
| Slurm | |
| AWS | |
| LLM evals (lm-eval, Daytona, Harbor, etc.) |
Languages
| German | |
| Native proficiency |
| English | |
| Professional proficiency (TOEFL iBT: 118/120) |
| French | |
| Basic proficiency (certified level B1, 4 years in high school) |
| Mandarin | |
| Basic proficiency (5 years in high school) |
| Spanish | |
| Basic proficiency (two semesters at Berliner Volkshochschulen) |
| Italian | |
| Basic proficiency (one semester at UC Berkeley) |
Interests
| Hobbies | ||||||||||||
| Biking | ||||||||||||
| Hiking | ||||||||||||
| Working Out | ||||||||||||
| Running | ||||||||||||
| Bouldering | ||||||||||||
| Tennis | ||||||||||||
| Badminton | ||||||||||||
| Kite Surfing | ||||||||||||
| Skiing | ||||||||||||
| Writing | ||||||||||||
| Chess | ||||||||||||
| Politics | |||
| German Green Party | |||
| European Horizons Chapter TU Berlin co-founder | |||
| Volunteering | |||
| Soup Kitchen Founder | |||
| Math Club Organizer | |||