Publications

PureML: a transparent NumPy-only deep learning framework for teaching and prototyping

Journal of Open Source Software · 2026

Introduces a compact NumPy-based framework for teaching and experimentation. It offers a middle ground between micrograd’s minimal autodiff engine and the complexity of general-purpose frameworks like PyTorch and TensorFlow or systems-oriented projects like tinygrad. Readers can follow the gradient rules in the source code, then use built-in layers, optimizers, and checkpointing to run complete training experiments.

SAWNERGY: Software architecture and methodology for dynamic residue-interaction networks and walk-based embeddings from molecular dynamics trajectories

Undergraduate honors thesis · Wesleyan University · 2026

Describes a pipeline that builds networks of interacting protein residues from molecular dynamics trajectories and uses random walks on those networks to learn residue embeddings for downstream analysis and machine learning.