PROJECT FEED

Ordinary differential equations (ODEs) are essential tools for capturing the time-dependent behavior of complex systems. Recent advances in machine learning have accelerated model discovery by deriving governing equations directly from observational data. While these so-called black-box models often achieve accurate predictions, they tend to overlook fundamental laws that are critical in chemical engineering applications. Here, we develop a hybrid framework that integrates first-principles-based feature engineering with data-driven techniques to uncover underlying physicochemical mechanisms. Our approach leverages genetic algorithms to identify multiple best-fitting solutions under user-defined constraints informed by a priori knowledge. Building on our prior success in identifying algebraic systems—both linear and nonlinear in parameters—we extend AI-DARWIN, our interpretable, mechanism-based modeling framework, to dynamic systems governed by ODEs. We demonstrate its robust performance across diverse domains, including atmospheric chemistry, cellular signaling, and electrochemistry, using synthetically generated sparse and noisy data.

The growing global demand for advanced electronics has accelerated the need for more efficient device fabrication methods. In particular, the fabrication of nanoarrays depends heavily on effective organization strategies, which traditionally fall into two categories: bottom-up approaches (assembling building blocks based on particle interactions) and top-down approaches (constructing arrays from bulk materials, such as through 3D printing). In this work, we focus on the bottom-up technique and present a game-theoretic framework to model the self-assembly of liquid metal particles. We conceptualize each particle as a utility-maximizing agent and simulate their competitive interactions to gain insight into the emergent macroscopic properties of the resulting nano-wire arrays. This predictive framework offers a computationally efficient alternative to advanced molecular simulations and provides critical information for the design of nanoscale materials. Our methodology extends our prior work in agent-based modeling of complex systems, where large populations of dynamical agents consume and dissipate energy, leading to emergent behaviors at the macroscopic scale.

LABS's interests are in developing and applying statistical, machine learning, and artificial intelligence (ML/AI) methodologies to “big data” in multi-omics and medical data for aging and diseases such as Alzheimer’s disease. We emphasize utilizing advanced ML/AI techniques and multi-omics data, including MRI, genetics, transcriptomics, and proteomics, to solve clinically relevant problems for precision medicine. The applicant will learn to process medical imaging data (e.g., brain MRI) using AI/ML and related computational tools for this project. All relevant directions can be pursued based on discussions, interests, and expertise.