MLOps with Dataiku

Deploying Projects to Production

The deployer is the central place where operators can manage versions of Dataiku projects and API deployments across their individual life cycles.

Manage code environment and infrastructure dependencies for both batch and real time scoring, and deploy bundles and API services across dev, test, and prod environments for a robust approach to updates.

Reliable Batch Operations

Dataiku automation nodes are dedicated production servers which execute scenarios for everyday production tasks like updating data, refreshing pipelines, and monitoring or retraining models based on a schedule or triggers.

With these dedicated execution servers, multiple AI projects run smoothly in a reliable and isolated production environment.

Real Time Results with API Services

Deliver answers on-demand with Dataiku API nodes —- elastic, highly available infrastructure that dynamically scales cloud resources to meet changing needs.

In just a few clicks, generate REST API endpoints for real-time model inference, code functions, SQL queries, and dataset lookups, leading to more downstream applications and processes powered by AI.

Monitoring & Drift Detection

Once AI projects are up and running in production, Dataiku monitors the pipeline to ensure all processes execute as planned and alerts operators if there are issues.

Model evaluation stores capture and visualize performance metrics to ensure that live models continue to deliver high quality results over time. When a model does degrade, built-in drift analysis helps operators detect and investigate potential data, performance, or prediction drift to inform next steps.

Model Retraining and Comparisons

Production models periodically need to be updated based on newer data or shifting conditions. Teams may either manually refactor a model or set up automated retraining based on a schedule or specific triggers, such as significant data or performance drift.

With comprehensive model comparisons in Dataiku, data scientists and ML operators perform champion/challenger analysis on candidate models to make informed decisions about the best model to deploy in production.

CI/CD with APIs for DevOps

Robust APIs enable IT and ML operators to programmatically perform Dataiku operations from external orchestration systems and incorporate MLOps tasks into existing data workflows. Dataiku integrates with the tools that DevOps teams already use, like Jenkins, GitLabCI, Travis CI, or Azure Pipelines.

Model Stress Tests and Auto-Documentation

With a series of stress tests simulating real world data quality issues, ML operators reduce risk by assessing model robustness and behavior under adverse conditions, prior to deployment.

Automatically-generated, customizable documentation for models and pipelines helps teams retain critical project context for reproducibility and compliance purposes while simultaneously reducing the burden of manual documentation.