Vertex AI is a unified platform that supports various technologies and tools on GCP to build an ML project from end to end.

Traditional challenges of building ML are:

• Handling large quantities of data
• Determining the right machine learning model to train the data
• Harnessing the required amount of computing power

There are also production challenges:

• Scalability
• Monitoring
• Continuous integration, delivery and training

Finally there are ease of use challenges:

• tools require advanced coding skills
• focus is taken away from model configuration
• there’s no unified workflow
• finding tools is difficult

Vertex AI is Google’s solution to these challenges.

Vertex AI provides end to end ML pipeline:

1. Upload data from different sources (Cloud storage, BigQuery, locally)
2. Create features (processed data to be input into the model), and share them with others through the feature store
3. Training and hyperparams tuning
4. Deployment and model monitoring

Vertex AI is also a unified platform that provides both Predictive AI and Generative AI.

• Predictive AI allows for sales forecasting and classification
• Generative AI enables the creation of multimodal content

AutoML and Custom training are also both on Vertex AI

One convenient feature is that data scientists can write SQL with Workbench on Vertex AI to seamlessly connect BigQuery to Vertex AI

Vertex AI is:

• Seamless: smooth user experience
• Scalable: scale compute and storage automatically
• Sustainable: artifacts and features can be shared
• Speedy: models have 80% fewer lines of code than competitors