Research
Research directions at the intersection of machine learning, computer vision, medical AI, scientific computing, and robotics.
Uncertainty-aware learning and reliable AI
This direction studies calibration, uncertainty estimation, out-of-distribution behavior, and robust evaluation for machine learning systems used in high-stakes settings.
Focus: trustworthy prediction, distribution shift, error analysis, and evaluation protocols for real-world deployment.
Typical artifacts: publications, benchmarks, evaluation code, and reproducible experiments.
Physics-informed neural networks and AI for scientific computing
This line of work combines machine learning with physical models, variational principles, and numerical methods for forward and inverse problems in mechanics and fluid dynamics.
Focus: PINNs, neural surrogates, non-Newtonian flow modeling, and computational methods guided by domain knowledge.
Typical artifacts: research papers, simulation code, notebooks, and comparative numerical studies.
AI in medicine, medical imaging, and radiogenomics
This research investigates clinically meaningful models for medical images and multimodal biomedical data, including tasks that connect imaging findings with pathology, genomics, or longitudinal clinical evidence.
Focus: medical image analysis, multimodal fusion, radiogenomics, clinical decision support, and interpretable prediction.
Typical artifacts: publications, curated datasets, analysis pipelines, and model implementations.
Robotics and vision-language-action systems
This direction explores how visual perception, language-conditioned reasoning, and action planning can be integrated in robotic and embodied AI systems.
Focus: perception-action loops, multimodal grounding, planning under uncertainty, and interactive robotic systems.
Typical artifacts: papers, prototype code, teaching materials, and demonstration videos.