Expedia Upgrades Ranking Models to Keras 3: 30% Faster Training, 70% Lower Latency

fchollet · x · 2026-08-12

Expedia recently migrated its ranking models to a state-of-the-art Keras 3 setup. The upgrade resulted in a 30% increase in training speed and a massive 70% reduction in inference latency.

Keras creator François Chollet highlighted the writeup as a great source of technical insights for serving large-scale ranking and recommendation models at ultra-low latency.

Related event: Expedia Migrates to Keras 3, Slashing Inference Latency by 70%(2 posts)→

Original post →

More from Infra

Infra channel →