MLOps Workflows and Best Practices on Google Cloud Platform with Scotiabank's Hoda Abokhadra

This talk by Hoda Abokhadra, Analytics Devops Engineer at Scotiabank, is optimal for Data Scientists and ML engineers who want to apply DevOps principles to ML systems (MLOps). That means practicing automation, monitoring, testing, releasing, deployment and infrastructure management.

The session will be divided into two parts. The first covers an overview of the MLOps lifecycle. It introduces MLOps processes and capabilities and why they’re important for ML-based systems.
This will include the MLOps lifecycle:
ML development.
Continuous training.
Model deployment.
Prediction serving.
Continuous monitoring.
Data and model management.

The second part is a deep dive into the MLOps best practices which are important for optimizing the ML tools and services in GCP like Vertex AI and BigQuery.
This will be the following:
Vertex AI workbench.
Storing ML resources and artifacts.
Using Vertex AI SDK for Python.
Preparing training data.
Storing data in BigQuery.
Using hyper-parameter tuning for model predictions.
Using feature attributions.
Using Dataflow for data processing.

This talk was originally delivered at Arize:Observe 2023, a conference on the intersection of large language models, generative AI, and machine learning observability in the era of LLMops.

Connect with Hoda on LinkedIn:   / hoda-abokhadra  
Get updates from Arize on future events: https://arize.com/community/
Get certified in ML observability: https://courses.arize.com