Federico Milani

Head of SW Development, Data, and AI @ Compremium AG

About Me

What I love most about my work at Compremium AG is knowing that the software we build ends up helping real patients and the clinicians who care for them. After years in research, it’s a great feeling to see ideas leave the lab and become something people actually use. Today I lead our Software Development, Data, and AI team. I spend a lot of time with clinicians figuring out what they really need, then turning that into medical software and applied AI that works within the strict rules of healthcare. That’s a hard puzzle, and I enjoy every bit of it. Whenever I can, I still get my hands dirty with prototypes.

I hold a PhD in Computer Science from Politecnico di Milano, with a research background in computer vision and deep learning.

I speak Italian and English fluently, get by in Spanish, and have basic German (A2). Outside work I love all kinds of sports, especially anything that gets me outdoors for a breath of fresh air while exploring Switzerland’s beautiful nature.

Experience

Compremium AG

https://compremium.ch

Head of SW Development, Data, and AI

August 2023 - Present

Politecnico di Milano

https://www.polimi.it/en/

Research Fellow

February 2019 - January 2023

Department of Electronics Information and Bioengineering (DEIB)

Researched on Deep Learning architectures and Computer Vision techniques for cultural heritage and remote sensing data. Worked with state-of-the-art image classification, object detection and instance segmentation models.

  • Led the ArtDL project, obtained state-of-the-art results on iconography identification through Deep Learning and Computer Vision techniques with potential use in interactive museums applications and improvement of online collections.
  • Actively took part in H2020 projects for the processing and analysis of remote sensing data, drastically reducing manual photointerpretation time and human errors in sensitive use cases (e.g., illegal landfills identification, drone search and rescue).
  • Co-authored 10+ publications in international journals and conferences in the Deep Learning and Computer Vision field.
  • Co-supervised 10+ MSc students with thesis focused on Deep Learning, Computer Vision, and Machine Learning. More than 50% of the thesis led to an international publication.
  • Assisted and tutored BSc and MSc courses for multiple academic years (2019 to 2023): Web Technologies (5 cfu, 300+ students), Data Bases 2 (5 cfu, 300+ students).

Education

Politecnico di Milano

Ph.D. in Computer Science

November 2019 - March 2023

Department of Electronics Information and Bioengineering (DEIB)

Researching on Deep Learning (DL) architectures and Computer Vision (CV) techniques for the analysis of Cultural Heritage data to support complex iconography studies and the enrichment of public online collections. Currently working on ArtDL.

Thesis title: “Analysis of Cultural Heritage Data for Complex Iconography Studies”

Research focus: Artificial Intelligence, Computer Vision, Image Classification, Object Detection, Data Analysis

Politecnico di Milano

MSc in Computer Science and Engineering

October 2016 - December 2018

MSc track on Ambient Intelligence and Data Engineering

Master thesis on the feasibility of Deep Learning (DL) architectures for landform extraction with the final aim to enrich Volunteered Geographic Information (VGI) systems, such as OpenStreetMap (OSM).

Thesis title: “Learning to find mountains”

Projects

Python, PyTorch, Tensorflow, Jupyter Notebook, Numpy, Pandas, Scikit-learn, OpenCV, Pillow

Created and released a public data set of 43k paintings, implemented and evaluated Deep Learning (DL) models for Christian iconography localization. Acquired expertise in data preparation and analysis, Artificial Intelligence, Computer Vision and data science Python libraries.

Python, Jupyter Notebook, Docker, Pandas, Plotly, Numpy

Contributed to the development of an innovative open-source framework assisting Machine Learning (ML) applications in the stages of data preparation, model evaluation, and black-box error diagnosis. ODIN is entirely developed following Python and unit testing best practices.

Publications

C. Fasana, S. Pasini, F. Milani, P. Fraternali Weakly Supervised Object Detection for Remote Sensing Images: A Survey (2022) MDPI - Remote Sensing

F. Milani, N. O. Pinciroli Vago, P. Fraternali Proposals Generation for Weakly Supervised Object Detection in Artwork Images (2022) MDPI - Journal of Imaging

P. Fraternali, F. Milani, R. N. Torres and N. Zangrando Black-box Error Diagnosis in Deep Neural Networks for Computer Vision: a Survey of Tools (2022) Springer - Neural Computing and Applications

R. N. Torres, F. Milani and P. Fraternali ODIN: Pluggable Meta-annotations and Metrics for the Diagnosis of Classification and Localization (2021, October) in LOD 2021: Machine Learning, Optimization, and Data Science

N. O. Pinciroli Vago, F. Milani, P. Fraternali and R. da Silva Torres Comparing CAM Algorithms for the Identification of Salient Image Features in Iconography Artwork Analysis (2021) MDPI - Journal of Imaging

F. Milani and P. Fraternali A Data Set and a Convolutional Model for Iconography Classification in Paintings (2021) ACM - Journal on Computing and Cultural Heritage

R. N. Torres, F. Milani and P. Fraternali Algorithms for mountain peaks discovery: a comparison (2019, April) In Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing (pp. 667-674)

R. N. Torres, P. Fraternali, F. Milani and D. Frajberg Mountain summit detection with Deep Learning: evaluation and comparison with heuristic methods (2019) Applied Geomatics, 1-22

R. N. Torres, P. Fraternali, F. Milani and D. Frajberg A Deep Learning model for identifying mountain summits in Digital Elevation Model data (2018, September) In 2018 IEEE First International Conference on Artificial Intelligence and Knowledge Engineering (AIKE) (pp. 212-217)