Francisco J. Peña‘s path from postdoctoral researcher at Digital Futures to assistant professor at KTH Royal Institute of Technology is a testament to how cutting-edge research can drive academic careers. During his time at Digital Futures, Francisco played a pivotal role in the Deep Wetlands project, applying machine learning and remote sensing to monitor vital ecosystems.
Today, as the head of the Learning Beyond Supervision Lab, he continues to push the boundaries of AI in healthcare and environmental applications. This article explores how his experience at Digital Futures shaped his research approach and supported his transition to a leadership role in academia.
Hi Francisco, first of all, how did you end up as a Digital Futures postdoc?
One of the things I love most about life is how unexpected it can be. My postdoc project at Digital Futures happened thanks to tennis. While working on large language models during my postdoc at KTH, I received feedback that my grant proposal would be stronger if it were multidisciplinary. I had just moved to Stockholm and didn’t know anyone outside computer science – except a professor from Stockholm University I played tennis with regularly. I asked about his research and whether artificial intelligence could help. He said yes, and that conversation led to a joint project that was later funded by Digital Futures. That’s how it all started.
How did your experience with the Deep Wetlands project at Digital Futures shape your approach to applying machine learning in real-world environmental challenges?
Working on the Deep Wetlands project was my first experience applying machine learning to a real-world problem with direct impact. Until then, I had worked with real data, but never in close collaboration with domain experts who needed practical solutions. The challenges were very different – noisy, incomplete data and no annotations. I spent about 80% of my time curating data and only 20% designing models, which was the opposite of my previous research. It completely reshaped how I approach real-world machine learning.
What were the key skills or knowledge areas you developed during your postdoc that have been most valuable in your current role as an assistant professor?
The most valuable skill I developed was learning to communicate and collaborate with researchers outside computer science. Explaining machine learning concepts and learning from experts in hydrology and remote sensing was a great experience. Writing interdisciplinary papers challenged me to be clear and precise. I also significantly improved my grant writing skills, which I now rely on daily to secure funding for new projects.
Can you describe how the interdisciplinary collaboration between KTH and Stockholm University influenced your research perspective?
The collaboration completely changed how I view research. I realized that improving performance isn’t about using bigger models – it’s about integrating domain knowledge. My approach now is to deeply understand the problem first, then design machine learning models tailored to that specific context. Since working with Stockholm University, I no longer apply generic models to new data; I build models specifically designed for the data and the problem at hand.
What inspired the creation of the Learning Beyond Supervision Lab, and how does its mission connect with your work at Digital Futures?
The Learning Beyond Supervision Lab was created to lower the barriers to using AI. The two main challenges are the need for annotated data and high computational resources. Our lab focuses on removing the first barrier – developing methods that work without labeled data. This connects directly to my work at Digital Futures, where we trained state-of-the-art wetland detection models without any annotations and made them accessible to environmental researchers. By solving the data problem, we also reduced the computational burden – our models can now be trained on a single GPU and run on a standard laptop.
Looking back, how did the mentorship and collaborative environment at Digital Futures support your transition to a faculty position at KTH?
Digital Futures played a key role in my transition to a faculty position at KTH. It gave me the opportunity to lead my first independent project and collaborate with researchers beyond computer science. That experience was a crucial step in my career, leading to a postdoc at Karolinska Institute and eventually to my current role at KTH. I’m proud to have done postdocs at all three universities of the Stockholm Trio – and even prouder to continue collaborating with all of them today.
If you’re working on an interesting problem and have a lot of data but no annotations, let’s talk. That’s exactly the kind of challenge we love to tackle in the Learning Beyond Supervision Lab. I’m always open to new collaborations – especially those that push the boundaries of what AI can do in the real world.
Text: Johanna Gavefalk
