Found his calling

(interview from 2026)

Dagmar Möbius

At school, he wasn’t exactly a star student in English, according to his language teacher. She never would have thought he’d go on to build a career in the U.S. But Dr. Hendrik Strobelt has been living and working in America since 2013 and is currently conducting research on artificial intelligence at IBM.

“At 18, I had a very local view of the world,” says Hendrik Strobelt. Born and raised in Zwickau, the tech-savvy young man decided in 1997 to study computer science at TUD Dresden University of Technology. His father had already studied Civil Engineering there, and his mother had studied the same subject in Leipzig. “Back then, I didn’t even consider going abroad.” At the time, he knew nothing about TUD’s excellent reputation or its international standing.

Turning a passion into a profession

“With an ABC electronics kit available in the GDR, you could even build a long-wave radio,” recalls Hendrik Strobelt. Sometimes the then 13-year-old was allowed to use the computer in his parents’ company in the evenings. Among other things, they worked with structural analysis programs. “The PC was sacred,” he laughs. “At first, my father knew more, but that changed quickly.” His computer science teacher at school also played a major role in his growing interest in the subject. “Once, some TVs had fallen off a truck somewhere; we then built small computers out of the leftover parts in class.”

As a teenager, he found the 1990s to be very lively and formative, “because you were no longer punished for freely expressing your opinion.” During the German reunification period, he attended a pilot class at Käthe-Kollwitz-Gymnasium. “There was a school club where we could buy a beer in 9^th grade,” he recalls. But equally: “The school still had coal heating, so two people had to shovel coal.”

© J. Lennerz

Prevented data breach and influential teachers

In 2007, Hendrik Strobelt successfully completed his studies at TUD, where he was also active on the Student Council. “In academic self-governance, I learned a great deal about dynamics in meetings. That still helps me today,” he says. He can hardly use the material taught back then anymore, since things in computer science change rapidly. What has remained, however, are fundamental, essential principles: “Don’t be afraid of the new, abstraction as a core skill for computer science, and a holistic view of problems.”

Three lecturers had a lasting influence on him. Professor Michael Schröder supervised his thesis - he valued Schröder’s experience in both business and academia. He learned a great deal about university policy and data security from Prof. Andreas Pfitzmann (ƚ) , with whom he also had really good conversations. When it comes to data security, Hendrik Strobelt recalls a funny incident: “Due to an error in the computer center, a file that was not supposed to be public had been made public.” Meaningless to laypeople, but “a challenge” for the computer scientist, because if you knew how, you could extract the exam results from the file containing secret numbers. The Data Protection Officer, upon learning of this, is said not to have been amused. Hendrik Strobelt followed Professor Oliver Deussen to Konstanz to earn his doctorate under his supervision in 2012, focusing on the visualization of large document collections.

New York – Boston – Cambridge

© archive H. Strobelt

After completing his doctoral thesis, Dr. Hendrik Strobelt moved to the United States. First to New York University, then to Harvard University in Boston from 2014 to 2017. Since 2017, the scientist has been working at IBM Research in Cambridge, Massachusetts. He is also a member of the MIT-IBM Watson AI Lab. The lab conducts AI research and develops algorithms that advance the economy and society. “My company,” says Dr. Hendrik Strobelt when he talks about IBM. The 110-year-old company may not be as glamorous as Meta and is perhaps a bit underrated, but it is a “very mature company.” According to him, it is one of the few independent research institutions where scientists enjoy a great deal of freedom in their research.

Current AI Research – Don’t use a sledgehammer to crack a nut

For his specific work, this means researching how to extract useful insights from AI models. One of the questions that interests him is how artificial intelligence makes decisions. One goal, for example, is to break down tasks for large AI models into smaller programs that can be run on a laptop. “Breaking large tasks down into smaller ones is a fundamental principle of computer science,” explains Dr. Hendrik Strobelt. He illustrates: “Today, small AI models consist of about seven billion parameters. What is interesting is that when AI models are used in traditional computer programs, probabilities suddenly come into play. Example: 1 + 1 = 2. Earlier models could certainly suggest the solution 2.5. This usually doesn’t happen anymore; the models are simply too good for that. But how do you rewrite code in such a way that probabilities are reduced and the result can be verified?” Or: “How do I use a model that knows what it’s supposed to do?”

© archive H. Strobelt

This work is also so crucial because, in many applications, large AI agents waste resources. Another aspect is preventing attacks on artificial intelligence models. “Don’t use a sledgehammer to crack a nut,” as Dr. Hendrik Strobelt puts it. “Boring for academics, highly interesting for practitioners.”

For a current project, he is collaborating with Berlin's Charité hospital. The project involves preliminary studies on how to automate information extraction from doctors' notes for cancer treatments. This very well correlates with his interest in biomedical topics. Another issue that is on his mind besides his work: “How to collect health data for research in Germany in order to make progress.”

Having worked in various fields such as electronics, data engineering, visualization, and artificial intelligence, Dr. Hendrik Strobelt believes he “found his calling.”

Contact:
Dr. Hendrik Strobelt (PhD)
Senior Research Scientist at IBM Research 
Explainability Lead at MIT-IBM Watson AI Lab
Visiting Researcher at MIT

Email:

Web