Netflix Tech news

The latest AI and developer news from Netflix Tech, with short summaries.

Netflix tied together models, features, and experiments into a single graph with its **Metadata Service**.

·2·2026.05.05 01:00

Netflix routed model serving traffic by purpose using Switchboard.

·1·2026.05.02 06:00

Netflix attached the FilmLight API to MPS to extend camera file processing to the cloud.

·2026.04.25 00:00

Netflix has built a BOC- and TOC-centered operations system to keep pace with live scaling.

·2026.04.18 00:00

Netflix used LLM judges and agents to automatically evaluate synopsis quality in line with human standards.

·2026.04.11 01:00

Netflix reduced duplicate queries from rolling Druid dashboards using interval-aware caching.

·1·2026.04.07 07:00

Netflix built real-time multimodal video search with a Cassandra-Kafka-Elasticsearch pipeline.

·2026.04.04 09:00

Netflix switched all Live events from CBR to VBR, maintaining quality while reducing traffic.

·2026.04.03 06:00

To match its scale of 300M+ members and 190+ countries, Netflix centralized its localization analytics.

·2026.03.07 00:00

Netflix reduced CPU usage in recommendation score calculation through batching, flat buffers, and the JDK Vector API.

·2026.03.03 10:00

The container scaling bottleneck wasn't containerd — it was the global mount lock on modern CPUs.

·2026.03.01 07:00

Netflix strengthened shot-level content understanding with MediaFM, which combines audio, video, and text.

·2026.02.24 03:00

Netflix built a flexible LLM Post-Training Framework on top of Ray, PyTorch, and vLLM.

·2026.02.13 17:00

Netflix built a self-service automated migration to move around 400 PostgreSQL clusters to Aurora PostgreSQL.

·2026.02.12 23:00

Netflix is leveraging LLMs to evolve its existing structured query-based Graph Search into natural language search.

·2026.01.27 04:00