Sitemap
A list of all the posts and pages found on the site. For you robots out there, here is an XML version available for digesting as well.Pages
Posts
No posts found. Please check back later.
Portfolio
No portfolio found. Please check back later.
Projects
RheoNet: A Physics-Inspired Neural Network for Predicting the Viscosity of Complex Fluids
May 2023
This project aims to develop a physics-inspired machine learning model for predicting the viscosity of complex fluids and to use it to aid in the formulation design process.
autoHSP: An Autonomous Experimentation Framework for Determining Hansen Solubility Parameters via Active Learning
July 2023
This aim of this project is to leverage machine learning and automated lab technologies to autonomously determine Hansen Solubility Parameters (HSPs) in a closed-loop fashion.
FormulationAI: A Hierarchical Machine Learning (ML) Framework for Predicting Emulsion Stability
October 2023
This project aims to develop a hierarchical machine learning model for predicting the stability of emulsions and to use it to develop sustainable formulations with greener chemicals.
ChemGuide: A Chemistry-Inspired Diffusion Model with Non-Differentiable Guidance
May 2024
This project aims to develop a chemistry-inspired diffusion model with non-differentiable guidance for generating novel molecules with desired properties while enhancing molecular stability.
Publications
Yttrium–Sodium Halides as Promising Solid-State Electrolytes with High Ionic Conductivity and Stability for Na-Ion Batteries
Published in The Journal of Physical Chemistry Letters, 2020
This paper is about demonstrating a type of Yttrium–Sodium Halide-based materials as promising solid-state electrolytes for Na-ion batteries.
Recommended citation: Qie, Y.; Wang, S.; Fu, S.; Xie, H.; Sun, Q.; Jena, P. Yttrium–Sodium Halides as Promising Solid-State Electrolytes with High Ionic Conductivity and Stability for Na-Ion Batteries. J. Phys. Chem. Lett. 2020, 11 (9), 3376–3383. 🔗
Download Paper
Chemistry-Inspired Diffusion with Non-Differentiable Guidance
Published in ICLR'25, 2025
This paper is about leveraging domain knowledge from quantum chemistry as a non-differentiable oracle to guide an unconditional diffusion model in generating more stable 3D molecular structures.
Recommended citation: Shen, Y.*; Zhang, C.*; Fu, S.*; Zhou, C.; Washburn, N.; Póczos, B. Chemistry-Inspired Diffusion with Non-Differentiable Guidance. The Thirteenth International Conference on Learning Representations. ICLR, 2025. 🔗
Download Paper
(under review) Predictive Modeling of Emulsion Stability via Hierarchical Machine Learning
Under review, July 2025
Stay tuned for updates on this exciting publication!
Under review: Roguski, M.*; Fu, S.*; Walker, L.; Tilton, R.; Washburn, N.; Cochran, B.; Stone, C.; Barker, M.; Jamadagni, S.; Johnson, E. Predictive Modeling of Emulsion Stability via Hierarchical Machine Learning. Under review, 2025.
(under review) Autonomous Determination of Hansen Solubility Parameters via Active Learning
Under review, August 2025
Stay tuned for updates on this exciting publication!
Under review: Fu, S.; Wang, D.; Henderiks, H.; Assis, A.; Charron, J.; Washburn, N. Autonomous Determination of Hansen Solubility Parameters via Active Learning. Under review, 2025.
Download Paper
(manuscript ready) RheoNet: A Physics-Inspired Neural Network for Predicting the Viscosity of Complex Fluids
Manuscript ready, August 2025
Stay tuned for updates on this exciting publication!
Manuscript ready: Fu, S.; Bruno, C.; Chua, A.; Póczos, B.; Washburn, N. RheoNet: A Physics-Inspired Neural Network for Predicting the Viscosity of Complex Fluids. Manuscript ready, 2025.
Presentations and Posters
Molecular Representation for Property Prediction: Wanders in the Chemical Space
December 2, 2022
Talk, Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania
From the early days of QSAR to the current deep learning models such as variational autoencoders and graph neural networks, molecular representation has been a central theme in cheminformatics. In this talk, I reviwed the history of molecular representation and discussed the current state-of-the-art models for property prediction.
Formulation AI for the Guided Viscosity Design of Complex Fluids
February 12, 2024
Poster, Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania
The design of complex fluids is a challenging task due to the high-dimensional formulation space. In this work, we developed a genetic algorithm-based and physics-informed formulation AI to predict the dynamic viscosity of non-Newtonian complex fluids. We showed how AI can be implemented to discover new viscosity models that are more related to real-life datasets than empirical models such as the Cross model. More information here.
Closed-loop and Autonomous: CMU Cloud Lab for Measuring Hansen Solubility Parameters
June 27, 2024
Talk, Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania
Autonomous labs can be a tranformative tool for the future of research to accelerate scientific discovery, but they are capital intensive and require a high level of expertise to design and operate. In this talk, I introduced our approach to use the CMU cloud lab as a proxy for designing autonomous workflows. Specifically, we demonstrated how the cloud lab can be used to measure the Hansen solubility parameters of a series of polymers in a closed-loop and autonomous fashion. Everything, including the design and selection of experiments, the execution of the experiments, and the analysis of the experiment findings, is done autonomously with no human intervention, except the neccessary human involvement in the cloud lab facility.
Teaching
No teaching found. Please check back later.