As data, analytics, and AI programs expand across organizations, various teams, systems, and methods can limit trust and success. New technology, like Generative AI chatbots and services, can further the problem by introducing another set of choices and technologies for teams to manage. Instead of considering these new technologies separate from other advanced analytics, data teams should leverage existing AI platforms as they bring Generative AI into their organization and projects.
A critical element of trust for enterprise AI projects is quality and consistency. This robustness created by a systematic approach helps everyone feel more confident in projects as there are fewer issues and concerns. One way to deliver high quality and consistency is to centralize the operations and governance of everything from data pipelines to the most sophisticated AI projects on a single platform. This gives data and operations teams a single way to manage projects, streamlining operations and improving governance.
Dataiku 12 includes new capabilities for data and IT teams to streamline MLOps and governance processes to deploy models faster, better manage production models, and improve model governance.
Model Overrides
A core principle of AI safety is keeping a human in the loop. Models don't always have the best or safest answer. New model overrides in Dataiku add a human layer of control over model outputs to enforce expected outcomes under specified conditions and guarantee compliance when results do not meet expectations.
Model Overrides in Dataiku 12
Improved Deployment and Monitoring
Dataiku 12 streamlines and automates model deployment and monitoring steps, saving time and minimizing errors. ML engineers and IT operators typically undertake complex manual reconfiguration and mapping tasks when deploying pipelines and models to production environments and building custom feedback loops.
This process can be slow and error-prone. New features in Dataiku 12 include a simplified monitoring setup, visualization of API deployments to projects, and new drift metrics, including KS, Chi2, and PSI. These new MLOps features reduce friction at handoff points during the MLOps lifecycle and automate more configuration settings, saving time and reducing risk.
New Governance Views
Dataiku 12 adds new governance views, including a Kanban view to enable teams to see all projects across all stages, from development to production. Teams can filter to narrow the focus to specific initiatives, regions, business functions, and internal sponsors and click on any project to see more detail. A configurable risk/value matrix also provides additional context to leaders.
New Kanban View Improves Project Visibility by Stage
Building trust in analytics and AI projects happens one project at a time when teams consistently deliver the expected results to the business. New AI technologies, like LLMs, are changing the landscape for data teams. Leaders must seriously consider how they implement these new technologies to continue to build trust. A critical component of this process is consistently applying best practices and processes across all aspects of project development, from data pipelines to model development, MLOps, and governance. Dataiku continues to be the place where teams can do it all.