Research
Research Directions
Our research is organized around three interconnected directions that follow a single idea: develop AI, make it safe and trustworthy, and apply it responsibly for society.
Direction 01 · Develop
AI Development
Building advanced intelligent systems
Our research explores how to build capable AI systems: models that generate, perceive, reason, and act. We approach model development with reliability and downstream use in mind from the start.
-
Generative AI
We investigate generative models, their mathematical foundations, and how their outputs can be controlled and evaluated.
-
Visual & Multimodal Intelligence
We study models that interpret visual information and connect it with language and other modalities.
-
Agentic & Embodied AI
We explore AI agents that plan and take actions, including agentic approaches to computer vision tasks.
Direction 02 · Make Safe
AI Safety & Trust
Making AI trustworthy, verifiable, and resilient
As AI systems become more capable, it becomes more important to understand, verify, and secure them. We investigate methods that help establish when AI systems, and the content they produce, can be trusted.
-
AI Forensics & Content Provenance
We investigate methods for detecting AI-generated content and for tracing the origin and history of digital media.
-
Verification, Evaluation & Explainability
We explore rigorous ways to evaluate model behavior and to make model decisions easier to inspect and understand.
-
Robustness, Security & Privacy
We study how AI systems respond to distribution shifts and adversarial manipulation, and how to protect the data they rely on.
Direction 03 · Apply
AI for Society
Advancing AI for responsible and beneficial real-world use
Our research explores how trustworthy AI methods can be applied to real-world problems, and what responsible use requires in practice.
-
Geospatial AI
We explore AI for remote sensing and geospatial data, where reliability directly affects downstream decisions.
-
Healthcare AI
We are interested in AI methods for health-related data, where safety, fairness, and interpretability are essential.
-
Responsible & Ethical AI
We consider fairness, accountability, and transparency as part of how AI systems are designed and used.
-
Critical Applications
We study the requirements of AI used in high-stakes settings, where errors carry significant consequences.
-
Sustainable AI
We explore ways to reduce the computational and environmental costs of developing and deploying AI.
Interested in these directions?
Students and collaborators are welcome to get in touch.