teaching/resources

Interactive tools, hands-on tutorials, open-source models, and university courses.

Interactive Demos & Simulators

UMA Playground

Interactive browser-based simulation demo from Meta AI

The UMA Playground allows researchers, students, and practitioners to interactively explore simulations with the Universal Model for Atoms (UMA) directly in the browser with zero installation. You can test diverse chemical systems spanning catalysis, inorganic materials, and molecules, run real-time structural relaxations, and visualize atomic trajectories.

Open UMA Playground

FAIRChem UMA Demo (Hugging Face Space)

Technical cloud demo for custom atomic structure simulations

Hosted on Hugging Face Spaces, this app enables users to upload custom atomic structure files (e.g. .cif, .xyz, .pdb) and configure simulation parameters such as temperature, charge, and simulation steps. It runs geometry optimizations and molecular dynamics (MD) trajectories in the cloud powered by UMA foundation models.

Open Hugging Face Demo

FAIRChem Tutorials & Open Source

The FAIRChem (Meta FAIR Chemistry) repository contains open-source implementations, trained model checkpoints, and comprehensive educational tutorials for machine learning in chemistry and materials science:

  • FAIRChem GitHub Repository: Core library housing architectures (UMA, EquiformerV2, GemNet, PaiNN, DimeNet++), data loaders, and training pipelines.
  • ASE (Atomic Simulation Environment) Integration: Hands-on guides showing how to use UMA and Open Catalyst models as drop-in ASE calculators for structure optimization, equation-of-state calculations, and transition state searches.
  • Molecular Dynamics (MD) Tutorials: Jupyter notebooks detailing how to set up, execute, and analyze NVE/NVT molecular dynamics simulations across diverse material and molecular systems.
  • Model Fine-Tuning & Custom Potentials: Step-by-step instructions on fine-tuning pretrained foundation models on bespoke DFT datasets, specific reaction pathways, and local cluster environments.
  • Open Datasets: Guides for downloading, preprocessing, and benchmarking with large-scale datasets including Open Catalyst (OC20, OC22, OC25), Open Materials 2024 (OMat24), Open Molecules 2025 (OMol25), Open Molecular Crystals (OMC25), and Open DAC (2023, 2025).

University Courses (Carnegie Mellon University)

During my faculty appointment at CMU, I developed and taught undergraduate and graduate courses in chemical engineering, computational modeling, and machine learning:

  • Mathematical Methods for Chemical Engineers (06-262): Core course for sophomore undergraduate students covering linear algebra, differential equations, optimization, and analytical problem-solving in chemical systems.
  • Numerical Methods and Machine Learning for Chemical Engineers (06-325): Required junior-level undergraduate course bridging scientific Python, numerical methods, optimization, and foundational machine learning techniques applied to chemical engineering problems.
  • Chemical and Reactive Systems (06-625): Graduate-level course on chemical reaction engineering, transport phenomena, heterogeneous catalysis kinetics, and advanced numerical modeling for MS and PhD students.