Which Machine Learning Classifiers are Best for Small Datasets…
A deep dive across 108 benchmark datasets to see what holds up when you only have 100–1,000 samples.
Blog
Long-form research notes and practical ML lessons from the field.
Underspecification in Machine Learning
How pipelines can return models with identical scores but radically different real-world behavior.
Building a Better Search Engine for Ai2
Lessons from using three years of search logs to upgrade Semantic Scholar's ranking system.
SHAP Values and Feature Variance
Why interpretability matters, and how SHAP plots change the conversation with stakeholders.