From 16 to 18 September 2026, the third International Reinforcement Learning Bootcamp brought 90 participants and speakers from 12 countries and more than 25 institutions to the University of Salzburg. Over three days, the free programme combined international keynotes, hands-on coding and exchange across the reinforcement learning community.
Hosted by the Smart Analytics and Reinforcement Learning team at the university’s IDA Lab, the bootcamp was organised in collaboration with AI Austria, the RL Community, JOANNEUM RESEARCH ROBOTICS and DIGITAL, DIH West and FFG.
The programme addressed a central challenge in reinforcement learning: translating research into systems that solve practical problems. Lectures and coding sessions took participants from foundational concepts to policy gradients, world models, humanoid robotics and agentic AI.
Four international keynote speakers contributed perspectives from research and application:
- Ronald Ortner, Montanuniversität Leoben, explored satisficing in reinforcement learning: approaches that seek sufficiently good outcomes.
- Michael Bowling, University of Alberta and Amii, examined what models should represent in model-based reinforcement learning.
- Verena Kain, CERN, discussed AI and the design and operation of particle accelerators.
- Elmar Rueckert, Montanuniversität Leoben, presented work on learning humanoid robot policies for industrial tasks.
Hands-on tutorials gave participants opportunities to implement algorithms and test their understanding. The programme included a maze-based coding exercise and a live Air Traffic Control tournament, connecting theoretical concepts with the challenges of training and evaluating agents. Participants trained their first RL agents, while the organisers also reported the start of new research collaborations.
The exchange continued beyond the lecture hall. Discussions over coffee, lunch and dinner created space to compare approaches, develop ideas and connect with others working on similar challenges. Participant feedback particularly highlighted the international speakers, organisation and welcoming community atmosphere.
AI Austria thanks the organising team, led by Simon Hirlaender and Olga Mironova, together with Andreas Windisch, Thomas Gallien, Marius-Constantin Dinu, Leander Grech and Léa Keller, as well as the local support team, speakers and participants who made the event possible.
The learning continues beyond Salzburg: recordings of all three days, presentation slides, tutorial notebooks and environments are openly available, allowing participants and the wider community to revisit the sessions and work through the exercises at their own pace.
Planning is already underway for the fourth International Reinforcement Learning Bootcamp in Salzburg in 2027. Drawing on participant feedback, the organisers aim to introduce clearer beginner and advanced tracks, more opportunities to write code and additional industry case studies.
In the meantime, the community stays connected through RL Coffee, a monthly virtual meetup on the first Friday of each month for discussing research, papers and new ideas.
Explore the bootcamp and continue learning: