Research in AI and engineering

Engineering design

Physics-informed surrogate models · differentiable optimization

We explore physics-informed machine learning and numerical optimization for multidisciplinary design. The aim is to make complex design studies more tractable while checking model predictions against suitable simulation and test references. Current directions include surrogate modeling, reduced-order models, and coupled design problems.

Diagnostics and sustainment

Prognostics and health management · state of health · remaining useful life

Our diagnostics research examines how battery and power-system data can reveal degradation and inform maintenance. We study state-of-health and remaining-life estimation, with attention to uncertainty, representative operating conditions, and the maintainer's workflow.

Engineering decisions

Decision assurance · model applicability · traceable engineering evidence

We study how to keep engineering recommendations connected to the requirements, assumptions, model versions, and evidence that support them. When those inputs change, the goal is to identify what needs to be checked again and preserve the engineer's approval authority.

AI infrastructure

Research depends on infrastructure that can run experiments, connect models, and retain useful records. Meso-AI-Gateway provides a common point for model access, policy, and usage tracking in a customer's Google Cloud environment.