I created cleanlab open-source in grad school at MIT used by thousands of data scientists across hundreds of Fortune-500 companies to automatically improve any ML model by automatically identifying most kinds of issues in any ML dataset (data errors, label errors, outliers, overlapping classes, regression, multi-label, multi-class, named-entity recognition, etc). cleanlab works with any ML model and every major ML and deep learning library (tensorflow, pytorch, xgboost, fasttext, huggingface, etc).

Specialize in learning with noisy labels, confident learning, and weakly supervised learning.

Website: https://www.curtisnorthcutt.com/

Recent work: Confident Learning: Finding Label Errors in Datasets and Learning with Noisy Labels