Back office ops · Production

From Facts & Metrics to Media Machine Learning: Evolving the Data Engineering Function at Netflix

The problem

Netflix's traditional data engineering focused on structured tables for metrics, dashboards, and statistical modeling, but as studio and content production scaled, media data — multi-modal, unstructured, and massive — required a fundamentally different approach that existing pipelines could not provide.

Workflow diagram · grounded in source

Media asset ingestion from AMP

integration

“All data for this phase comes from AMP, our internal asset management system and annotation store”

Asset standardization

validation

“Standardizing media assets across modalities (video, images, audio, text) to ensure consistency and quality for ML applications”

ML model inference on media

ai_action

“media tables — structured datasets that not only capture traditional metadata, but also include the outputs of advanced ML models”

Media Table storage in Data Lake

output

“We have partnered with our data platform team to pilot integrating LanceDB into our Big Data Platform”

API and UI access for researchers

output

“An pythonic interface that will provide programmatic access to the Media Table, supporting both interactive exploration and automated workflows”

Reported outcome

Netflix formalized a new Media Data Engineering specialization and built the Media Data Lake to provide centralized, standardized, scalable access to media assets and ML-derived features, enabling richer ML models, faster experimentation, and new AI-powered features.

Reported metrics

ML model qualityRicher, more accurate ML models

Experimentation and productization speedFaster experimentation and productization

Operational efficiency in title launchincrease operational efficiency in title launch preparation and ratings

Reported stack

LanceDBAMP

Source

https://netflixtechblog.com/from-facts-metrics-to-media-machine-learning-evolving-the-data-engineering-function-at-netflix-6dcc91058d8d

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Netflix formalized a new Media Data Engineering specialization and built the Media Data Lake to provide centralized, standardized, scalable access to media assets and ML-derived features, enabling richer ML models, fa…

What tools did this team use?

LanceDB, AMP.

What results were reported?

ML model quality: Richer, more accurate ML models; Experimentation and productization speed: Faster experimentation and productization; Operational efficiency in title launch: increase operational efficiency in title launch preparation and ratings (source-reported, not independently verified).

How is this back office ops AI workflow structured?

Media asset ingestion from AMP → Asset standardization → ML model inference on media → Media Table storage in Data Lake → API and UI access for researchers.