Research Projects
Impacts of Multiple Climate Stressors and Extreme Events on Biodiversity in Climate Change and Climate Intervention Scenarios: A Big Data Approach
Mentor: Dr Phoebe Zarnetske
Description: Climate change is increasing the frequency and intensity of extreme events, such as heatwaves, droughts, floods, and storms, resulting in cascading impacts on organisms and biodiversity. This project aims to develop a computational framework to identify and analyze the spatio-temporal impacts of multiple climate stressors and extreme events under both climate change and climate intervention (e.g., Stratospheric Aerosol Injection) scenarios. By leveraging large-scale climate datasets and advanced computational tools, including machine learning, spatial statistics, and time-series analysis, students will process and analyze extensive climate data to map where and when multiple stressors co-occur during extreme events. This approach will identify spatiotemporal "hot spots" of ecological vulnerability and provide critical insights into how climate interventions might change the frequency, intensity, and interactions of extreme events. The findings will be instrumental in guiding conservation efforts and informing strategic decisions to address the impact of climate change.
Generating Drug Molecules Inside a Protein Binding Pocket
Mentor: Dr Alex Dickson
Description: Diffusion models are generative machine learning models that have shown much recent success in creating data such as text (LLMs), images (DALL-E), audio, video and more. Diffusion models have also been applied to generate conformations of atoms representing proteins and drug molecules, however there are concerns about generalizability to new chemical structures and interactions that aren’t observed in the training set. In this project the student researcher will apply a recently developed method to integrate diffusion models into physics-based molecular dynamics simulations and assess its performance on a number of targets. This project integrates computation with aspects of physics, chemistry, molecular biology and pharmacology. Students from all of these backgrounds (and more) are encouraged to apply!
SEE-Insight Lab – Advancing Scientific Discovery with AI
Mentor: Dr Dirk Colbry
Description: The SEE-Insight lab focuses on developing large-scale, advanced artificial intelligence tools to accelerate scientific discovery. Our current research includes creating innovative solutions and tools for annotating scientific images, fine-tuning large language models (LLMs), exploring active learning techniques, developing meta metrics for evaluating AI outputs, and addressing challenges associated with limited LLM prompt inputs. We welcome students at all levels to join our team and contribute to cutting-edge projects, gain valuable skills, and engage in impactful research in a collaborative, supportive, and dynamic environment.
Modeling Spontaneous Transitions in Brain Dynamics
Mentor: Dr. Mengsen Zhang
Description: The human brain is never at rest, even when you are doing nothing! Instead, the brain spontaneously switches between different dynamic states when no external task is imposed. In dynamical systems frameworks, such spontaneous transitions are named “metastability” (read more in this paper here: https://t.co/wiBuG3ut1I). However, appropriate data analysis tools are still missing to capture the complexity of metastable brain dynamics fully. In this project, the student researcher will first construct a dynamical system model to capture the metastable feature of brain dynamics and then apply topology data analysis to differentiate the dynamic regimes of the model brain.
Classical Simulation of Noisy, Large-Scale Quantum Circuits
Mentor: Dr Ryan LaRose
Description: Classically simulating quantum circuits is important theoretically to understand the power of quantum computers and practically to benchmark algorithms and error correction. New classical simulation techniques can even challenge or refute claims of quantum advantage [https://arxiv.org/abs/2412.14703]. In this project we seek to emulate Google and IBM quantum computers by incorporating the noise models of these devices into methods for noisy simulation. Our methods build on the BGLS simulation algorithm [https://dl.acm.org/doi/10.1145/3624062.3624215] and a technique to exploit noisy channels to reduce simulation complexity. The goal of the project is to implement and optimize these methods on HPC resources to classically simulate noisy quantum circuits on the order of 100s of qubits and 1000s of quantum operations.
Impact of Social Network Structure on Opinion Dynamics
Mentor: Dr Lianzhang Bao
Description: In recent years, opinion dynamics has become a prominent topic in network science, drawing significant attention from economists, mathematicians, and physicists. The binary agreement model is a classical framework used to study opinion dynamics, where individuals on a network interact with only one of their neighbors at each time step. This model identifies a tipping point (10%) at which a committed minority can dominate the entire social network in the long run. However, in real-world scenarios, individuals often interact with more than one neighbor at a time, making the binary agreement model less applicable in such contexts. Inspired by the division of labor observed in social insects (e.g., ants and honeybees), we introduce a new variable to measure the strength of each individual's opinion. This extension allows us to adapt the binary agreement model to accommodate multiple-agent interactions at each time step. In this project, we aim to investigate whether the number of interactions can play a role analogous to the tipping point observed in the classical binary agreement model.
Understanding Reaction Mechanisms using Ab-Initio Molecular Dynamics
Mentor: Dr Samik Bose
Description: The mechanism of chemical reactions is characterized via a combination of experimental techniques such as spectroscopy, isotope labeling, kinetic assays, which are both time and cost intensive. One can also computationally model a chemical reaction using the laws of quantum mechanics to view the reaction mechanism and provide information about activation energies and transition states. Computational modeling of reactions is tremendously helpful in the context of catalysis in chemistry and biology as it can explore the avenues to increase catalytic activity, without explicitly carrying out the reactions in an experimental laboratory setting. Ab-initio molecular dynamics simulation is the tool that allows one to carry out chemical reactions in computers. In this project, the student will learn the basic concepts of ab-initio MD and help implement an enhanced sampling ab-initio MD simulation method to study well known catalytic reactions and their mechanisms. Students applying for this project should have experience in python programming.
Efficient and Scalable Path-Planning for Optimal Control and Differential Games
Mentor: Dr Christian Parkinson
Description: Partial differential equations (PDE) based models for optimal control and differential games provide a fully interpretable alternative to black box machine learning algorithms. However, classical numerical methods for PDE suffer from the curse of dimensionality, making them infeasible for complex, high-dimensional systems. In this project, we will design numerical methods for PDE based on variational formulations which scale well to high dimensional problems and can thus be applied to complex systems in (nearly) real-time. We will be particularly interested in path-planning problems in high-dimensional spaces and involving multiple agents. This project will be accessible to students who have passed the calculus sequence, as well as introductory differential equations and linear algebra, and have some prior coding experience.
PHYSICS-INFORMED MODELING OF COMPLEX DYNAMIC SYSTEMS FROM DATA
Mentor: Dr Huan Lei
The far-reaching impact of the analytical frameworks of modern physics on both our understanding of complex physical systems and the rapid development of advanced technologies in sectors like aerospace, materials science, and bioengineering cannot be overstated. However, many challenges remain. For example, we struggle to accurately model the nanoscale heat transfer relevant to CPU design or the dynamics of protein folding. For such complex systems, there are no analytical models available to accurately characterize their evolution. This project is designed to introduce recent developments in machine-learning to the development of equations governing complex, multiphysics systems while ensuring that these data-driven models strictly preserve all the physical symmetries and constraints of the system.
Particle-in-Cell Models to Study Instabilities in Fusion Plasmas
Mentor: Dr Andrew Christlieb
Nuclear fusion is a path to clean renewable energy; however, the burning plasma is not easily confined as instabilities arise at the micro scale and grow to the macro scale. Ideally, a deep understanding of these instabilities provides a pathway to the threshold for nuclear burn. In this project we will develop computational tools to study the instabilities with a primarily focus on Particle-in-Cell models.