Research¶
My research sits at the intersection of AI safety, dependability, explainability, and engineering methods for safety-critical systems. I am especially interested in assurance workflows that make AI systems inspectable, monitorable, and useful in real-world critical domains.
Research Themes¶
Responsible and Trustworthy AI¶
I develop techniques for assessing and improving the trustworthiness of AI systems, with emphasis on responsible deployment, safety monitoring, fairness, uncertainty, and evidence that can support decisions in critical settings.
LLM and Multimodal AI Safety¶
Recent work focuses on domain-specific safety monitoring for large language models, multimodal AI alignment, document-heavy decision workflows, and human-agent collaboration for specifying safety and alignment objectives.
Explainable AI¶
I work on robust, model-agnostic explainability methods that help users understand why models behave as they do. This includes SMILE/XWhy and explanation workflows for safety-critical ML applications.
Dependability of Autonomous Systems¶
I study model-based safety, runtime assurance, fault diagnosis, and digital-twin techniques for autonomous systems, drones, multi-robot systems, and industrial AI applications.
Methods¶
| Method family | Typical use |
|---|---|
| Statistical safety monitoring | Detecting unsafe or uncertain ML behavior at runtime. |
| Explainable AI | Producing interpretable evidence for model behavior. |
| Model-based safety engineering | Connecting architectures, hazards, faults, and runtime assurance. |
| Digital twins | Evaluating behavior and dependability under realistic operating conditions. |
| Human-agent workflows | Supporting collaborative specification, review, and assurance of AI behavior. |
Academic Service¶
- Associate Editor for the International Symposium on Model-Based Safety and Assessment.
- Associate Editor for IEEE/RSJ International Conference on Intelligent Robots and Systems.
- Guest Associate Editor for special issues on AI safety, trustworthy neurosymbolic AI, and manufacturing technology.
- Reviewer for IEEE, Springer, Elsevier, and safety-critical systems venues.