// MODEL OPTIMIZATION AND PROMPT SYNTAX TERM

Canary Deployment

Canary deployment means releasing a new model version to a small group of users first to test it before a wider rollout.

TECHNICAL DEFINITION

Canary deployment is a progressive rollout strategy where a new model version is deployed to a small subset of production traffic or users, allowing for real-time monitoring and quick rollback if issues arise, minimizing impact.

BACKGROUND

Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making. It is a field of research in engineering, mathematics and computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximize their chances of achieving defined goals.

READ MORE ON WIKIPEDIA

SYNONYMS & ALIASES

  • Phased rollout
  • gradual release
  • canary release

USAGE NOTE

Canary deployment reduces the risk associated with deploying new machine learning models.

DEVELOPERS

Organizations developing technology related to Canary Deployment.

  • LaunchDarkly

    Provides a feature flagging platform that enables developers to perform canary deployments and controlled rollouts of new AI features, prompt variations, or model versions by toggling them on for a subset of users.

  • Split.io

    Offers a feature delivery and experimentation platform that allows for robust canary deployments and A/B testing of AI models, prompt engineering changes, and AI-driven application components in production environments.

  • Seldon Technologies

    Specializes in MLOps and model serving, providing the Seldon Core platform which enables advanced deployment strategies like canary releases and A/B testing for machine learning models, ensuring safe and controlled rollouts of new AI capabilities.

  • Databricks

    Through its MLflow platform and MLOps capabilities, Databricks helps manage the lifecycle of machine learning models, supporting practices like canary deployments for new model versions or prompt-tuning adjustments in production.

  • Weights & Biases

    Offers an MLOps platform that facilitates experiment tracking, model evaluation, and deployment management, which is crucial for implementing controlled rollouts and canary deployments of AI models and their associated prompt strategies.

  • Amazon Web Services (AWS SageMaker)

    Provides a comprehensive managed service for machine learning, including model deployment endpoints that support A/B testing and traffic splitting, enabling canary deployments for different versions of AI models or prompt engineering strategies.

  • Google Cloud (Vertex AI)

    Offers a unified platform for ML development and deployment, featuring model serving capabilities that allow for traffic management and gradual rollouts, essential for canary deployments of AI models and prompt engineering experiments.

  • Vercel

    For applications integrating AI models or prompt-generated content into user-facing web experiences, Vercel's deployment and Edge Config features can be used to perform canary releases of new UI/UX elements powered by different AI outputs or prompt designs.

RELATED TERMS IN MLOPS & DEPLOYMENT