From notebook to production: Serving JAX at scale
Jax Model serving Production ml Nvidia gpu Tensorflow serving Aot compilation Jax2tf Google cloud Jax.export Inference optimization Deep learning deployment
This video teaches developers how to deploy JAX models to production at scale, covering AOT compilation to lock down input shapes for predictable latency, jax.export for packaging portable execution graphs, and jax2tf conversion for TensorFlow Serving integration. It is designed for ML engineers and data scientists who need to serve JAX models in production environments with low latency requirements.
Related videos
Scale JAX models to multi-GPU systems
Shipping custom models at scale from fine-tuning to inference | BRK234
Emergent: How Google engineers actually ship AI (official trailer)
What's new for AI on GKE: Training, serving, and agents
From pilot to production: Scaling Gemini with proven practices
Gemma 4 production stack: Model Armor, ADK Agents, Tracing
Automate Google Cloud with Cloud CLI Remote MCP Server
Launch your portfolio website with AI
Launch your to-do web app with AI
Launch your web-based video game with AI
Build a real-time voice AI agent with Gemini Live API
