2017: Training outstanding doctoral students at IMPRS-IS – Interview with Lukas Schott

This month, we’re looking back to 2017 when Cyber Valley’s graduate school, the International Max Planck Research School for Intelligent Systems (IMPRS-IS), was founded. IMPRS-IS trains outstanding doctoral students from around the world and supports them in their PhD research so they can use their AI expertise to build a better future in Europe.

Lukas Schott is an alumnus of IMPRS-IS. He joined in 2017 to work with Matthias Bethge and Wieland Brendel. His PhD project investigated how to make AI ‘see’ more robustly so it could work safely in unpredictable environments. Lukas previously worked at Bosch (a member of the Cyber Valley Corporate Network) and has recently started at Aleph Alpha (a member of the Cyber Valley Start-up Network) with the aim of training sovereign LLMs in Europe. In the following interview, he reflects on what he learned during his PhD and how his training at IMPRS-IS has impacted his career. 

When you look back at your PhD, is there a moment that stands out as a defining highlight? What made it meaningful? 

The defining highlights were the collective 'Eureka' moments within the research group. Uncovering the hidden mechanisms behind AI behavior and gaining a deeper understanding of its inner workings made the long hours absolutely worth it. 

From your perspective, what is unique about IMPRS-IS? What distinguishes it from other doctoral training programs? 

I consider IMPRS-IS to be the leading AI graduate school in Germany. It distinguishes itself through an exceptional network that connects the Universities of Tübingen and Stuttgart with two Max Planck Institutes and leading industry partners, leading to exceptional AI research with high international impact. 

When you joined IMPRS-IS, what were your expectations for the program and your PhD journey? Were there any surprises along the way? 

I anticipated a purely academic focus, so the flexibility toward industry was a welcome surprise. Being able to intern at Google X [now: X, the Moonshot Factory] and Amazon Web Services without stalling my research was a unique advantage that really broadened my perspective. 

In what ways did IMPRS-IS shape your development and how do those experiences influence your work today?

It allowed me to peek behind the curtain and grasp the inner workings of AI systems. This experience refined my problem-solving skills and taught me to remain humble, always questioning assumptions since things often don't work the way we think they do. 

How does your work (both during your PhD and current) address real-world challenges and what kind of applications or long-term benefits could it have? 

My PhD work on robustness against corruptions (like rain or blur) directly addresses the safety of vision systems in the real world. My work on neural network compression solves the energy and hardware constraints of edge computing, allowing powerful Foundation Models to run locally on devices like automobiles. In my current job, we are training sovereign LLMs for Europe. 

IMPRS-IS is the graduate school of Cyber Valley, a network of leading research institutions, industry partners, and start-ups working in AI and robotics. How has being part of this network influenced your development, whether in your research, skills, or career direction? 

The strong reputation of the research community in Tübingen was a door-opener. It gave me the credibility to get a foot in the door for competitive international internships, which was crucial for my professional growth. 

To celebrate Cyber Valley's ten-year anniversary, we'll be sharing a variety of content across our website and social media channels reflecting on milestones from the last decade. Stay tuned for the next installments!