Tooling for agile modeling on large machine perception embedding databases.

Modular, scalable library to train ML models

Google Research

Recipes for reproducing Analysis-Ready & Cloud Optimized (ARCO) ERA5 datasets.

TabFM (Tabular Foundation Model) is a pretrained tabular foundation model developed by Google Research for tabular data regression and classification.

T2J-Bench is a benchmark for semantic-preserving codebase conversion.

DP-FTRL from "Practical and Private (Deep) Learning without Sampling or Shuffling" for centralized training.

CoReNet is a technique for joint multi-object 3D reconstruction from a single RGB image.