Uber Eats Cuts Search Latency by Half with Pipeline Overhaul
Uber Eats has overhauled its search pipeline, reducing the time users wait for initial search results to appear by 50%. The improvements included changing how latency is measured, optimizing candidate retrieval, splitting processing steps, reworking the advertising system, and updating underlying infrastructure. Uber reports Above-the-Fold search latency dropped by over 200 milliseconds after these changes, with no single major redesign, but many incremental gains across the full stack.
- Above-the-Fold latency reduced by more than 200 ms
- Candidate retrieval and hydration were streamlined
- Advertising path and data access were redesigned
- Infrastructure changes improved garbage collection and encoding
- No single fix; gains came from broad incremental changes
Sources covering this
In this story
More in Enterprise
Databricks introduces batch vector search for massive datasets
Databricks has added an engine-native vector search feature designed to handle massive batch jobs—instead of just fast real-time queries.
Google Cloud launches Modernize portfolio with AI tools
Google Cloud rolled out Modernize, a suite that bundles its migration, modernization, and infrastructure optimization tools into one…
Windows update causes crashes in some legacy apps
Microsoft has acknowledged that the Windows 11 KB5124010 preview update, released in September 2026, is causing certain games and apps…
Windows 11 is dropping legacy printer driver support
Microsoft is phasing out support for old-school printer drivers in Windows 11, shifting to its own built-in print system based on the…