GCP Machine: Cricket Video Analysis to Predicting Coal Mine Equipment Failures (Cloud Next '19)

Опубликовано 11 апреля 2019, 19:06

Two very different companies - Teck Resources, one of the world’s largest mining companies vs English and Wales Cricket Board, a major sports and entertainment organization. Two very different ML domains - predictive maintenance vs sports analytics. Two very different classes of data to work with - structured time-series IoT data vs unstructured video streams. Two very different levels of existing GCP adoption - from nothing in the cloud to already in the cloud. And yet, there are many common elements in building a successful machine learning solution.

Diving into the details, this session demonstrates how to take an idea to a working production ML solution while peeking into the technology along the way including:

Turning business needs into supervised learning problems and understanding the importance of focus. Examples of real-world challenges with labeling data - something that “synthetic” ML problems (hello MNIST) or Kaggle competitions simply don’t have.

Working with structured and unstructured data on leveraging Cloud Storage, DataFlow, Pub/Sub, BigQuery, Bigtable, Cloud Composer, Kubernetes Engine and more.

Leveraging various machine learning frameworks from Scikit Learn to TensorFlow to pre-build AI APIs as well as “semi-trained” models with specific examples of using Cloud ML Engine, Vision API, Video Intelligence API, AutoML.

Designing and implementing ML solutions as microservices with a synchronous or asynchronous API layer using Cloud ML Engine, AppEngine, Kubernetes Engine, Cloud Endpoints.

Using foundational cloud infrastructure elements from Cloud Build for continuous integration to Stackdriver for operational monitoring and insights.

GCP Machine Learning Applications → bit.ly/2WQC1Fw

Watch more:
Next '19 DevOps & SRE Sessions here → bit.ly/Next19DevOpsSRE
Next ‘19 All Sessions playlist → bit.ly/Next19AllSessions

Subscribe to the GCP Channel → bit.ly/GCloudPlatform


Speaker(s): Alexander Gorbachev

Session ID: MLAI234 product: Cloud - General; fullname: Alexander Gorbachev; event: Google Cloud Next 2019;