About me

I am a PhD student at the xAILab Bamberg, under the supervision of Christian Ledig. My research focuses on developing machine learning models that are robust, efficient, and privacy-aware, particularly for deployment in real-world healthcare settings. I work on improving generalization under distribution shifts while reducing the dependence on large training datasets and high computational resources.

Previously, I worked on Bayesian uncertainty estimation for dedicated breast-CT scans at the Advanced X-ray Tomographic Imaging lab, under the supervision of Marco Caballo. I’ve also interned at NATO and AIKO, where I worked on anomaly detection in underwater communications and visual odometry in planetary-like environments, respectively.

Outside of research, I'm a slow but enthusiastic runner. Feel free to follow my progress on Strava!

News

• [03/2025] One paper accepted at TMLR!
• [10/2024] Won the Best Paper Award Runner Up at ADSMI @ MICCAI 2024!
• [09/2024] Attended the M2L Summer School
• [07/2024] One paper accepted at BMVC 2024!
• [07/2024] One paper accepted at ADSMI @ MICCAI 2024!
• [06/2024] One collaborative paper accepted at MICCAI 2024!
• [04/2023] Started my PhD at the xAILab Bamberg
• [09/2022] Started a research internship at NATO
• [04/2022] Started a research internship at AXTI Lab

Show more

Selected papers

• An Embedding is Worth a Thousand Noisy Labels (TMLR, 2025)
  Francesco Di Salvo, Sebastian Doerrich, Ines Rieger, Christian Ledig
  ARXIV GITHUB
• Rethinking model prototyping through the medmnist+ dataset collection (Scientific Reports, 2025)
  Sebastian Doerrich, Francesco Di Salvo, Julius Brockmann, Christian Ledig
  ARXIV GITHUB
• Privacy-preserving datasets by capturing feature distributions with Conditional VAEs (BMVC, 2024)
  Francesco Di Salvo, David Tafler, Sebastian Doerrich, Christian Ledig
  ARXIV GITHUB
• MedMNIST-C: Comprehensive benchmark and improved classifier robustness by simulating realistic image corruptions (MICCAI-W, 2024)
  Francesco Di Salvo, Sebastian Doerrich, Christian Ledig
  ARXIV GITHUB
• Self-supervised Vision Transformer are Scalable Generative Models for Domain Generalization (MICCAI, 2024)
  Sebastian Doerrich, Francesco Di Salvo, Julius Brockmann, Christian Ledig
  ARXIV GITHUB

Service

• Reviewer for MICCAI 2025
  

Teaching

• Mathematics for Machine Learning (2023 - 2025)
  
• Robust Machine Learning (2023 - 2025)
  
• Thesis supervision (5x Master's, 1x Bachelor's)
  Contributed to one published outcome, and one thesis award.

Awards

• Third place at the DMEA Sparks Awards for a supervised BSc thesis
• Best Paper Award Runner Up at ADSMI @ MICCAI 2024

Side projects

• The Illustrated Machine Learning (628 stars and 20+ daily users, as of May 2025)
  Website containing a collection of visualizations and explanations of various machine learning concepts.

Email

francesco.disalvo99[at]gmail.com