In 2023, the world of data is an ecosystem. Organizations must weave through a landscape of systems, products, and platforms to achieve their business targets.
Many companies struggle with AI models that lack explainability or have no practical use in the business. It is crucial to make these models explainable and user-friendly for non-technical users. Additionally, centralizing these models in a single repository would make managing and maintaining them easier. Dataiku looks to solve all these challenges, even if you previously created and deployed these models externally.
Building off our MLFlow integration, we have extended the “external models” feature. You can now utilize your existing AWS SageMaker, Microsoft AzureML, or Google Vertex AI models in Dataiku.
By integrating external models, you can leverage the benefits of traditional Dataiku models for cloud models deployed externally. You'll get enhanced explainability and evaluation and can enable business users to easily access your organization's cloud models with an easy-to-use UI for broader AI adoption. Additionally, a comprehensive view of all your deployed models is in your organization’s Dataiku Govern model registry.
Dataiku provides a centerpiece for the MLOps lifecycle.
Explain and Validate External Models
Connecting external models to Dataiku’s platform can significantly enhance traditional cloud models' visibility, monitoring, and benchmarking. For example, you can take advantage of explainability features, such as universal feature importance and what-if analysis, to better understand how your cloud models behave and dynamically validate them.
Use what-if analysis on AWS SageMaker models.
When you use an external model in Dataiku, you can easily monitor your cloud models' performance with an evaluate recipe and compare them with other alternatives like models designed in Dataiku or with MLFlow. Additionally, you can utilize automated scenarios, metrics, and checks to create benchmarking logic for the performance of your Dataiku and cloud models.
Use Evaluate recipes to validate external models or compare them to Dataiku-designed models.
Democratize Cloud AI Models for All
Previously, non-technical users faced challenges in utilizing code-based models. However, using external models in Dataiku, business users can now leverage deployed models and streamline the process to run predictions and get outputs. Coupling a scoring recipe with an external model allows non-technical Dataiku users to obtain results from a code-based cloud model efficiently.
Score external models to receive quick predictions from cloud models.
Centralize MLOps for Models Deployed Anywhere
As mentioned earlier, Dataiku aims to be your organization's focal point for MLOps, providing a central platform to govern and monitor every model, regardless of its origin. This complete oversight becomes particularly evident when using external models in the Dataiku Govern model registry.
External models can be viewed and managed from the model registry in Dataiku Govern.
After declaring an external model in Dataiku, you can govern it like any other native model, which makes it easier to manage and monitor all the models in one place.
Dataiku is committed to offering a comprehensive AI solution for your organization, as it is already the go-to platform for all AI-related tasks, thanks to the seamless integration of MLFlow and the ability to incorporate external models. Soon, Dataiku models will be deployable anywhere (Cloud, Snowflake, Databricks, etc.), enabling seamless integration with external systems.