Research focuses on training and evaluating vision‑language agents for robust tool use, with large‑scale multimodal datasets and stepwise benchmarks. Applies computer vision in healthcare and environmental sensing; earlier work includes Android malware detection.