Introduction
A short setup for what this blog is for, what I plan to write about, and how often I expect to post.
Read articleBlog
Short write-ups on papers, methods, and ideas I want to revisit.
Data and foundational MLIPs
A rewritten reflection on data scaling, reference surface consistency, and why foundation MLIPs are only as coherent as the labels they learn from.
Read articleWhen Foundation Potentials Learn the Same Chemistry
Notes on platonic representations, shared chemical manifolds, and diagnostics for foundation machine learning interatomic potentials.
Read articleLearning the density first
A reflection on image super-resolution inspired electron density prediction, why learning real-space fields matters, and where density models can fit into materials discovery workflows.
Read articleWhen crystal prediction learns from its mistakes
A reflection on DAO, Siamese foundation models for crystal structure prediction, and why unstable structures can become useful training data for materials discovery.
Read articleThe dataset is the PES
A reflection on MatPES, why potential energy surface datasets need deliberate sampling, and why data quality can matter more than raw scale for foundation MLIPs.
Read articleWhen electrons keep score
A reflection on FlowER, conservation-aware reaction prediction, and why chemistry models should predict electron movement rather than unconstrained product strings.
Read article