Hi, I’m Hunter! 👋
I’m a machine learning engineer exploring NLP, adversarial ML, and AI applications in science. I share what I’m learning, document projects I’m working on, and discuss my research. Learn more about me →
Modern PyTorch Techniques for VAEs: A Comprehensive Tutorial
This tutorial bridges the gap in existing VAE literature by integrating modern PyTorch functionalities like torch.distributions and dataclasses for more efficient, cleaner code. Aimed at advancing understanding and application of VAEs with the latest PyTorch features.
Can You Hear the Shape of a Molecule? (Part Three)
After unsupervised clustering failed for larger molecules, we test whether supervised learning can extract hidden patterns from Coulomb matrix eigenvalues to classify alkane constitutional isomers.
Sarcasm Detection with Transformers: When Perfect Accuracy Isn't Perfect
This case study demonstrates how easy it is to accidentally train a domain classifier instead of a sarcasm detector. Using a pre-trained RoBERTa model, I achieve near-perfect accuracy on a benchmark dataset—only to discover the model learned to distinguish news sources rather than detect sarcasm.
Can You Hear the Shape of a Molecule? (Part Two)
Using clustering metrics to evaluate how well Coulomb matrix eigenvalues can separate alkane constitutional isomers without supervision.
Can You Hear the Shape of a Molecule?
Replicating a study that asks whether we can distinguish molecular shapes from their Coulomb matrix eigenvalues—a fundamental question about how much structural information these mathematical signatures contain.