Domino changes the game for AI and machine learning flows

Introducing Domino Flows from Domino Data Lab – the game-changing innovation that redefines AI and machine learning flows.

Domino Flows unlocks the power of AI and machine learning flows, transforming them into a critical element of your responsible AI strategy. AI and machine learning flows are composable or independent tasks that can be executed as a single process, such as tasks associated with transforming data or training a machine learning model.

AI and machine learning flows can be complex, often encompassing hundreds of steps. Designing and managing these flows can be challenging, and their complexity can make optimizing for performance, tracking, and reproducing results challenging. With Domino Flows, these tasks are simplified. For instance, Domino Flows automatically versions all aspects of even your most complex flows for instant reproducibility. It is built on top of Flyte, the leading open-source flow orchestration solution, and tightly integrated into the Domino Enterprise AI Platform.

Fully reproducible flows

While Flows are a common aspect of data science, Domino Flows takes this further. It's not just about integration. It's about full reproducibility. All results are tracked, versioned, and immutable, so you always have a permanent record of any work on the flow. Combined with Domino’s Reproducibility Engine, this makes flows and all data science work fully reproducible. Domino maintains accurate, immutable records and completely reproducible flows to meet even the most demanding regulatory and compliance needs.

Empower collaborative data science

With Domino Flows, you don't need to be a data science expert to manage complex flows. Domino Flows fosters collaboration and empowers data scientists by allowing them to leverage the full power of Domino’s collaborative data science platform. It enables the reuse of flows, capturing and sharing best practices. Flows can be templated and reused across projects. Subflows help you create nested flows that can be shared and run independently. Changing initial inputs to flows is as easy as changing a single parameter, meaning data scientists can quickly iterate on different data sets simply by pointing at a new data set and re-running their validated flows, which can be a massive time saver.

Dependency-aware flows

With Domino Flows, data scientists can work in the programming language of their choice at any given step in the flow because Flows is dependency-aware. It allows teams to design and execute AI and machine learning flows across any data source, infrastructure, or environment on-premises, cloud, or hybrid using Python, R, SAS, or any other language. This flexibility allows data science teams to innovate at various steps in a flow, using their favorite tools without creating dependencies. Because steps are containerized, teams can focus on their work without worrying about potential dependencies and conflicts, leading to exponential gains in data science productivity. Domino Flows can also run dynamically and include conditional logic, so flows run more efficiently.

Efficient, flexible AI/ML flows

Efficiency is paramount when executing compute-intensive tasks in an extensive, complex AI or machine learning flow. Domino Flows addresses these concerns by providing ultimate flexibility with self-service access to the best infrastructure for the job. Access to auto-scaling infrastructure makes processing significantly more efficient. Data scientists can use the DAG to inspect the entire flow quickly and more readily identify and fix bottlenecks, as well as use timelines to pinpoint steps that can be optimized in their flows.

Cost-effective flow processing

Domino Flows makes it easy and significantly reduces the costs of re-running flows. For example, Domino Flows uses caching of outputs from each step, eliminating the need to recalculate the entire flow whenever a change is made or an error is encountered. When a job fails, you can easily view details and re-run a flow from the point of failure. Cached outputs mean you don’t have to re-run costly computations, which enables more experimentation at a lower cost.

Secure, governed flows

Domino Flows shares the comprehensive governance and security framework that is a hallmark of the Domino Enterprise AI Platform, allowing it to serve the most highly regulated, highly secure industries and use cases. With Domino Flows, decision-makers gain unparalleled insight into every action taken within an AI/ML flow, allowing you to be confident your flows are compliant. Domino Flows tracks the movement and transformation of data throughout the lifecycle of your flow, including data access audit trails, so you know who accessed what data, from which data sources, and when. Flows ensures granular assignment of access rights, so only users with appropriate permissions can see and interact with flows or sub-flows in a project. Managers are alerted if any changes are made to a flow for full transparency.

Maintain data sovereignty

Domino Flows has separate control and data planes, which helps you comply with data sovereignty requirements by choosing which Domino Nexus data plane each step is executed on. Flows are backed by the comprehensive security inherent in the Domino platform. You can rest assured your flows are robust, secure, and governed to move them from development to production quickly.

Flows for any industry

Unlike other flow solutions, Domino Flows is:

• Comprehensive and flexible to evolve with your needs.
• Efficient and robust so even your most complex flows run smoothly and cost-effectively.
• Responsible and compliant with full traceability and reproducibility.

Whether you're in finance, life sciences, public sector, or any industry, Domino Flows doesn't just manage your machine learning flows – it masters them. Step into the future of AI and machine learning with Domino Flows and experience the total value of reproducible, efficient, cost-effective, and governed data science flows.

Next steps

See Domino Flows in action in this video.