Painless Machine Learning in Production

Explore a comprehensive talk from the EuroPython 2020 conference that delves into the challenges of managing machine learning models in production environments. Learn about the infrastructure and tooling used by a small team of data scientists and engineers to orchestrate the entire machine learning model lifecycle. Discover insights on balancing off-the-shelf solutions with custom development, prioritizing developer ergonomics, and empowering data scientists to deploy their work without relying on a dedicated MLOps team. Gain valuable knowledge about various technologies and techniques, including AWS SageMaker, Airflow, Docker, Cookiecutter, property-based testing, jsonschema, linting, Slack integration, model artifacts and diagnostics, automated deployments and rollbacks, healthchecks, autoscaling, and DBT. Walk away with practical insights to immediately apply to your own ML systems and infrastructure, and understand how to minimize engineering and ops overhead across data science teams of any size and composition.

Introduction
What is this talk about
About the company
Agenda
premises
Machine Learning Life Cycle
Machine Learning Data
The Early Web
ML Ops
Model Development
Features
Architecture
Principles
Airflow