SURF research bootcamp 2025

Speaker: Chris van Run - Research Software Engineer Radboudumc

Location: Experience Centre - Radboudumc

Language: English

Required:

• Laptop
• Minimum Python programming

Prior knowledge: You need to make a few small preparations. You can find those here: https://srb2025.grand-challenge.org/tutorial-preparation/

Learning Objectives: At the end of the session, you will understand how Grand Challenge manages the execution and evaluation of algorithms in a reproducible way. You have a clear idea of how submissions are processed and you can understand and design your own benchmarking workflows.

----------

In this practical workshop for research software engineers, we cover the core components of the platform Grand-Challenge.org. Using challenges, we will explore algorithms and evaluations and participants will de-bug, modify and test their own benchmarks.

Whether you develop models, organise and support challenges or are just curious about how large-scale benchmarking is automated, this workshop provides a compact and practical introduction to the infrastructure behind AI validation.

Grand-Challenge.org is an open platform for hosting AI benchmarks in medical imaging.

Speaker: Pieter Zeilstra - Researcher and lecturer in blending data science & software engineering at Saxion and the Ambient Intelligence Research Group. 

Location: Experience Centre - Radboudumc

Language: English

Required: Laptop

Prior knowledge: Some programming knowledge of python

Learning Objectives: At the end of this session, you will understand the basic principles of MLOps. You will have insight into how this field connects to the broader ambition of DEMAND and you will have an idea of how to use this to combine knowledge and translate it into practical tools that are directly applicable in organisations.

Maximum number of participants: 15

----------

The same thing happens in many organisations: someone trains a model that seems promising in a trial run. But as soon as it has to run in the real world, the problems arise. Data changes, systems do not connect properly and experiments are difficult to repeat. The result? The model gets stuck in a proof-of-concept and fails to deliver lasting value.

This is exactly where MLOps offers a solution.It is all about robust pipelines, monitoring performance, smartly organising experiments ánd integrating models into existing systems and processes. This way, a model is not tested once, but becomes a sustainable and usable part of the organisation.

In this interactive workshop, you will get a clear overview of the most important concepts, challenges and tools within MLOps in two hours.Among other things, you will discover:

• How to automatically train machine learning models with AutoML;
• How to version control data with DVC;
• How to make experiments and results transparent with MLflow;
• How to bring all this together in a scalable ML pipeline.

With practical examples, we will show you how to translate ideas, research and prototypes into solutions that really keep working in practice.

This session is aimed at data scientists, researchers, software engineers and other professionals working with data and AI who want to make their work more impactful by getting models production-ready.

This workshop is part of the SPRONG DEMAND project, a collaboration between three universities of applied sciences (HAN, Saxion and Fontys) and 16 partners from the field. 

Speaker: Babajide Owoyele - Researcher Radboud University

Location: Experience Center - Radboudumc

Language: English 

Required:

• Basic computer literacy (comfortable navigating files and web applications)
• A laptop with internet access
• Ideally, Docker installed on your laptop for local setup (we can provide installation instructions beforehand)

For participants who cannot set up Docker locally, we will provide a Google Colab notebook as an alternative, ensuring everyone can participate fully regardless of their technical setup.

Participants may also bring their own research video data if they wish to discuss how MaskAnyone may support their own projects during the hands-on session. We have some illustrative videos that will be shared via drive for the course.

We'll guide participants through the entire process step-by-step, from installation to practical application, making it accessible even for those without technical backgrounds. 

Prior knowledge 
No programming knowledge is required for this workshop. The MaskAnyone toolkit is designed to be accessible to researchers from various disciplines through its user-friendly web interface.

Learning Objectives
By the end of this workshop, you will be able to:

1. Understand the ethical landscape - Recognize privacy challenges in audio-visual research data and the importance of de-identification
2. Master the MaskAnyone Toolkit - Install, Configure, and Operate the MaskAnyone Software for Video De-Identification.
3. Apply anonymisation techniques by selecting and combining appropriate algorithms (e.g., blurring, pixelation, contours, solid fill) based on research needs and computational resources.
4. Implement privacy-preserving workflows - Integrate MaskAnyone into existing research pipelines to enhance data reproducibility while maintaining ethical standards.
5. Evaluate de-identification quality - Assess the effectiveness of different de-identification strategies and their impact on research utility.

Maximum number of participants: 20 
This range will allow for meaningful hands-on interaction while ensuring we can provide adequate support to all participants during the workshop.

----------

MaskAnyone is a modular toolkit designed to navigate privacy and ethical concerns of sharing audio-visual data in research. This hands-on workshop will introduce participants to a scalable, user-friendly solution for de-identifying individuals in video through subject masking, supporting multi-person behaviour tracking and analytics, and real-time bulk processing.

Participants will learn how to use this Docker-packaged web application to anonymize research videos while preserving data utility. The workshop covers practical implementation strategies for integrating privacy-preserving techniques into research workflows, addressing the growing need for ethical data management in social and behavioural sciences research.

The workshop is designed for researchers from various disciplines who work with audio-visual data and want to implement ethical data sharing practices in their work.

Speaker: Carsten Schelp - Consultant Scientific Compute Infrastructure at SURF and Yuliia Orlova - Consultant SURF Research Cloud at SURF

Location: Experience Center - Radboudumc

Language: English 

Required: Laptop with Google Chrome or Firefox, access to Wi-Fi (for instance Eduroam)

Prior knowledge: Experience with Linux commands is useful but not required.

Learning Objectives:

By the end of this workshop, you will be able to:

• Understand the concept of SURF Research Cloud

• Start your own working environment

• Use your own working environment

----------

Do you want to create and manage your own work environment and run powerful applications on it? SURF Research Cloud opens the door to cloud computing on different platforms.

This workshop explains cloud computing in general. The accompanying hands-on approach introduces you to (some of) the many features of SURF Research Cloud. You will learn how to create and manage your own virtual computing environment(s). By the end, you will feel comfortable enough to install your own applications and benefit from the power of cloud computing. During this training, your workload runs on the SURF HPC Cloud platform in Amsterdam.

An introduction to the SURF Research Cloud for anyone who is looking for a flexible and customized environment to run powerful applications.

Speaker: George A.K. van Voorn - Academic Tenure Tracker at Wageningen University & Research

Location: Experience Center - Radboudumc

Language: English

Prior knowledge: In principle no specific prior knowledge or experience is needed, except for some experience in developing (academic) model code.

Maximum number of participants: 24

----------

In this session the participants are introduced to a serious game as simulation of typical issues model or software developers may run into when working on real-life model development and application, e.g. for policy or commercial applications. We then use the results of the game for a discussion about, and introduction of, essential model and software quality requirements.