Skip to content

SYSTEM Cited by 1 source

Yelp ML Platform

Yelp's ML Platform is Yelp's internal framework for developing, storing, and deploying ML models across the company. Per Yelp's disclosures it is built around MLflow (experiment tracking / model registry) and MLeap (portable model serialization + JVM inference): ML engineers store MLeap-based XGBoost, neural-network, and other models in MLflow, then use an internal config manager to specify which models each downstream system should load.

Why it matters for system design

  • A model registry as the integration point. Downstream serving systems (e.g. Nrtsearch clusters) declare, via config, which model bundles to pull from MLflow at bootstrap — decoupling model iteration from serving-code deploys.
  • Compatibility as a design goal. A stated design goal for Yelp's Nrtsearch Inference Plugin was compatibility with the ML Platform, so existing models integrate without re-authoring.

Seen in

  • Yelp — ML based ranking using Nrtsearch (2026-05-11). The Nrtsearch Inference Plugin is designed to be compatible with the ML Platform; replica nodes load model bundles from MLflow at bootstrap and run MLeap-based inference in-process.

Stub page. Expand as dedicated sources arrive.

Last updated · 766 distilled / 2,225 read