This is 2hour introduction to Product AI on Google Cloud Platform. I will cover the basic components in Google’s latest AI service, AI Platform (Unified). I will walk the audience through two end-2-end production pipeline notebook, showing how AI Platform (Unified) is integrated into enterprise level production.
We will take a look at integrating AutoML, custom training for Tensorflow jobs, deployment to cloud instances, serving binaries, custom pre- and post- processing, auto-scaling, containers and debugging deployed models.
The two end-to-end pipelines we will discuss:
* AutoML Image Classification for online prediction
* Custom Training Raw Bytes (image) Classification
For each pipeline, we will deep dive to:
* Step by step sequences.
* Parameter choices.
* CSV and JSONL dataset (input) and prediction (output) formats.
* GPU and CPU compute and container selection.
* Single device, multi-device and multi-instance distributed training.
* Instance scaling for prediction
* Opinionated tips and best practices for integration.
Andrew Ferlitsch (Google)
Andrew is a machine learning expert at Google. he educates software engineers in machine learning and artificial intelligence. He is the creator of and oversees the development of the open source project Gap, which is a ML data engineering framework for computer vision. Andrew was formerly a principal research scientist at Sharp Corporation, working on imaging, energy, solar, teleconferencing, digital signage, and autonomous vehicles.