Technical Advancements
My technical research advancements focus on multimodal learning, foundation models, and agentic AI with explainability and interpretability as a central theme. More recently, I also focus on understanding and mitigating hallucinations in LLMs and grounding their knowledge in clinical evidence and guidelines.
Below are some of the selected publications for different technical themes. D. Mehta* corresponds to corresponding/first author.
Human Activity Monitoring & Action Recognition
Automated motion analysis and marker-less clinical assessment of movement disorders.
Vision-Language Models
Combining visual and language understanding for interpretable medical image analysis and clinical report generation.
Foundation Models & Agentic Systems
Large-scale medical foundation models and AI agents grounded in clinical evidence and guidelines for diverse healthcare tasks.
Interpretable and Concept-Based Learning
Explainable AI models where predictions are grounded in human-understandable visual and clinical concepts, enabling transparency and accountability for clinical adoption.
