Machine Learning driven Material Design

Discovering methodologies of how new materials can be found by combining traditional simulation techniques with machine learning methods is the newest research interest. I am very excited to find new ways to address years old challenges with new methods to enable new insights!

Generative BigSMILES: Cheminformatics for automated Simulations

Introducing generative BigSMILES (G-BigSMILES) – a groundbreaking enhancement to the BigSMILES notation tailored for generative workflows. G-BigSMILES streamlines the representation of intricate polymer ensembles, integrating key data like reactivity ratios and molecular weight distributions. Paired with its generative graph algorithm, it enables efficient molecule generation from specific polymer sets. This advanced tool sets a new foundation for automated polymer materials design, paving the way for robust machine learning techniques to grasp complex polymeric structures. By merging G-BigSMILES with cutting-edge machine learning, we're transforming property determination and automating in-silico polymer synthesis, poised to revolutionize polymer science and materials design.

Further details can be found here: Generative BigSMILES: An Extension for Polymer Informatics, Computer Simulations & ML/AI

CRIPT: A Community Databank for Polymers

The Community Resource for Innovation in Polymer Technology (CRIPT) provides a platform for people working in polymer science and engineering to capture and share data. CRIPT is led by a team at the Massachusetts Institute of Technology (MIT) along with collaborators in Academia, Industry, and Government, with support from the NSF.

I am a proud member of the CRIPT team, and have build the data model for simulation data and I am engaged in designing and realizing the Python SDK.

 

Learning Kinetics from Particle-Based Simulations

The defect relaxation in materials is one of these challenges. With my polymer background, I am focused on defects in the microphase separation of diblock copolymers. With machine learning, I was able to create a model, that could accelerate the simulation by training it with data from our particle based simulations.

Full details are here in this publication: Combining Particle-Based Simulations and Machine Learning to Understand Defect Kinetics in Thin Films of Symmetric Diblock Copolymers

Poster presentation for APS 2022 in Chicago: 3rd Prize for Poster Presentation

Previous
Previous

Mutliscale Entangled Polymer Dynamics

Next
Next

Block copolymer morphologies