Vincent Fortuin
Full Professor at Helmholtz AI
Munich, Bavaria, Germany
Summary
👤
Senior🎓
Top SchoolVincent Fortuin is a Full Professor and research group leader with 9 years of experience advancing Bayesian deep learning through better priors and more efficient inference. He leads the Efficient Learning and Probabilistic Inference for Science (ELPIS) group at Helmholtz AI and holds a professorship at Technische Universität Nürnberg, building bridges between academic research and applied science. His work spans top institutions—ETH Zürich, Cambridge, Google—and includes tangible open-source contributions to high-profile projects like Google’s uncertainty-baselines, where he implemented Posterior Networks and enhanced SNGP models for OOD evaluation. Known for combining theoretical rigor with engineering impact, he brings expertise in heteroscedastic modeling and practical uncertainty estimation for real-world systems. Trained originally in molecular life sciences and computational biology, he blends interdisciplinary insight with a deep technical focus on probabilistic modeling.
9 years of coding experience
2 years of employment as a software developer
Bachelor of Science (B.Sc.) Molecular Life Sciences, Bachelor of Science (B.Sc.) Molecular Life Sciences at University of HamburgDoctor of Science Machine Learning, Doctor of Science Machine Learning at ETH Zürich
Github Skills (9)
neural-network10
bayesian-methods10
machine-learning10
probabilistic-programming10
deep-learning10
tensorflow10
python10
data-science9
computer-vision8
Programming languages (6)
Github contributions (5)
google/uncertainty-baselines
Aug 2021 - Oct 2021
High-quality implementations of standard and SOTA methods on a variety of tasks.
Role in this project:
ML Engineer
ML EngineerContributions:7 commits in 1 month
Contributions summary:Vincent implemented and refactored models related to uncertainty estimation in deep learning. This included adding and modifying the Posterior Network model, which is a core component for the project's focus on Bayesian methods. The user further enabled OOD (Out-of-Distribution) evaluation functionality on the SNGP (Spectral-normalized Neural Gaussian Process) model and added the capability to test on a validation split. Additionally, the user implemented a heteroscedastic SNGP model.
implementationsstatisticsdata-sciencedeep-learningneural-networks
ratschlab/SOM-VAE
Jun 2018 - Aug 2020
TensorFlow implementation of the SOM-VAE model as described in https://arxiv.org/abs/1806.02199
Contributions:8 commits, 5 pushes, 10 comments in 2 years 1 month
information-theoryautoencoderarxivabsunsupervised-learning
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.