TECoSA has welcomed seven new postdoctoral researchers to the center. In this interview series, we will introduce the talented researchers joining our community and explore the innovative work they are pursuing. This interview features Zirong Chen.
Hi Zirong, what were you doing before signing up as a post doc at TECoSA/KTH?
I recently completed my PhD in Computer Science at Vanderbilt University, Nashville, TN, USA, where I worked on trustworthy AI and NLP for safety-critical applications. This research involved developing and deploying AI systems for emergency response in collaboration with the Metro Nashville Department of Emergency Communications.
What is your research focusing on?
My research focuses on trustworthy AI for safety-critical systems, particularly how we can make AI and large language model-based systems more reliable, explainable, and adaptable. I am especially interested in combining learning-based AI with formal reasoning and system-level safeguards so that increasingly capable AI models can be used responsibly in real-world critical applications.
What challenges does your research address/what problem does it solve?
Modern AI systems can be remarkably capable, but their behavior can also be uncertain and difficult to predict or explain. In safety-critical applications, a system being correct most of the time is simply not enough, so my research explores ways to detect uncertainty, impose meaningful constraints, and provide stronger assurances about how AI systems will behave when deployed in the real world.
What excites me most is the gap between making AI systems more capable and making them genuinely dependable, says Zirong Chen.
What makes this field particularly exciting to you?
What excites me most is the gap between making AI systems more capable and making them genuinely dependable. I have had the opportunity to see AI systems move from research prototypes into real emergency-response workflows, and that experience convinced me that the most interesting problems often begin when an AI model leaves the lab and has to work reliably with real people under real-world constraints.
What do you look forward to the most during this period as a Postdoc?
I am especially looking forward to working with researchers and industry partners across TECoSA and learning from perspectives beyond my previous work in emergency response. I hope to develop a broader research agenda around trustworthy AI that connects advances in AI with the rigorous safety and systems research at KTH.
Do you have a hidden talent that most people don’t know about?
I played competitive basketball as a power forward throughout high school and college, and before starting my PhD I also interned at the NBA (as a data scientist; unfortunately, not as a professional baller).
