My current interests are organized around two closely related research areas.
AI & multimodal security
How can visual and multimodal AI systems remain secure, robust, and auditable under adversarial inputs, weak evidence, and real-world uncertainty?
Security / robustness / adversarial and failure analysis / content integrity / trustworthy evaluation
Vision & multimodal intelligence
How can computer vision and vision-language systems preserve visual evidence, reason across modalities and time, and be evaluated for grounding rather than fluency alone?
Computer vision / VLMs / multimodal reasoning / memory / evaluation
Additional interests. Quantitative machine learning, interactive systems, and research software.
Some active research projects remain private during development and review; code, results, and reproducible artifacts will be released when appropriate.