OpenMLDB

Why we included this project

OpenMLDB is a machine learning database built around a simple problem: the features used to train a model should be the same ones the model sees at inference time. It stores and computes features so offline training and online serving read from one consistent source, and it handles the unglamorous plumbing around time-windowed aggregations and feature backfilling where data leakage usually sneaks in. Teams doing latency-sensitive scoring, such as fraud detection, risk scoring, and real-time recommendations, get a single place to define and serve features rather than maintaining separate batch and online stacks. And because feature logic is expressed in SQL, engineers who already know the language can get moving without picking up a new transformation framework.