IBM & Red Hat on AWS
Optimize AI governance with Amazon SageMaker and IBM watsonx.governance
In the rapidly evolving landscape of artificial intelligence and machine learning (AI/ML), governance has emerged as a critical imperative. As organizations increasingly rely on AI to drive business decisions and automate processes, ensuring enforcement of policies and standard practices has become a top priority. Amazon Web Services (AWS) and IBM have joined forces to provide an innovative AI governance integrated service, that help organizations build responsible AI products and meet their business, regulatory and compliance obligations.
On May 21st 2024, AWS and IBM announced an integrated product offering for AI/ML governance. IBM watsonx.governance now integrates with Amazon SageMaker, Amazon’s fully managed service to prepare, build, train, and deploy AI/ML models at scale. This integration provides customers with a simplified path to automate risk management and regulatory compliance for their AI/ML models and use cases.
Amazon SageMaker customers will have the ability to seamlessly share information like Amazon SageMaker Model Cards and Model Registry, with IBM watsonx.governance on AWS. This helps customers initiate comprehensive processes for assessing risk and adherence to corporate and regulatory policies, including the recently approved European Union Artificial Intelligence Act. This integration allows organizations to establish customizable risk assessment and model approval workflows, which can be triggered and tracked across multiple stakeholders, providing a complete model governance audit trail at every stage in both services.
Integrated Offering for Model Governance on Amazon SageMaker
With this integrated offering, customers gain access to highly scalable governance, risk, and compliance capabilities built to monitor and manage risk and compliance at scale:
1. Model risk governance: Map policies, metrics, and models using a centralized location to organize, document, and maintain an enterprise-wide view of your model inventory.
2. Operational risk management: Integrate risk and control assessments, internal and external loss events, key risk indicators, and issue/action plans within a single environment.
3. Regulatory change management: Combine software, process automation, data feeds, and expertise for a more complete, accurate, and timely view of your compliance risks.
Figure 1 below, demonstrates how the integration between IBM watsonx.governance and Amazon SageMaker enables the capabilities listed above:
Figure 1. IBM watsonx.governance and Amazon SageMaker integration flow.
For existing Amazon SageMaker customers with an inventory of models, the integration offers a seamless transition to enhanced governance capabilities. They can now benefit from comprehensive dashboards and customizable reports, with an option listed directly in their Amazon SageMaker console to activate IBM watsonx.governance.
Similarly, new Amazon SageMaker customers seeking to implement robust governance practices from the outset can leverage this integrated solution to test, validate, deploy, and monitor models across their entire lifecycle, ensuring responsible and ethical AI deployments.
GigaSpaces Technologies has chosen IBM watsonx.governance on AWS for AI governance
GigaSpaces is developing eRAG, a transformative solution that aims to change the way people interact with enterprise data through Generative AI. GigaSpace’s vision is to bring natural language interaction with structured data to a human level of accuracy.
As with any AI development project, ensuring responsible use is crucial. For GigaSpaces, this means making sure the eRAG text-to-SQL models are built and deployed in a way that’s secure, reliable, and compliant with regulations.
The new integration offering of IBM watsonx.governance and AWS SageMaker helps GigaSpaces monitor and manage risks throughout the AI lifecycle for eRAG models, ensuring responsible AI development practices.
“Our vision with eRAG, our enterprise RAG solution, is to empower organizations to make trustworthy business decisions when accessing structured enterprise data via natural language. Our collaboration with IBM and AWS is helping transform this vision into reality: We selected IBM watsonx.governance on AWS for its completeness and fit-for-purpose AI governance capabilities, as well as its ability to leverage the power of Amazon SageMaker to accelerate the ML lifecycle.” said Adi Paz, CEO of GigaSpaces Technologies.
Perform AI governance for Amazon SageMaker with IBM watsonx.governance
User-focused dashboards, reports, and automated collaborative tools provide invaluable insights into the state of AI use cases risk across the organization. Dashboards show AI/ML activities and how they align with your organization’s existing procedures, fostering transparency and accountability.
You will be able to see your Amazon SageMaker model inventory replicated on IBM watsonx.governance. From where you create model dashboards to visualize key model metrics, including models by departments, providers, risk levels, lifecycle phase of your models and among other details (figure 2).
Figure 2. Dashboard overview of key AI/ML model metrics, including risk levels.
You can evaluate new AI use-cases for your business with IBM watsonx.governance. Simply create a new Model Use-Case providing information such as use-case name, owner, purpose and description (figure 3).
Figure 3. Create and evaluate potential new AI use-cases for your business.
You can request risk assessment for your use-cases. There are a variety of assessments you can choose from, each evaluating a specific domain like potential regulatory risk. Based on the evaluation result, the use-case is given a risk posture that can be used to decide if the use-case is ready to be implemented.
Once a use-case is created, you can initiate your approval workflows and route it to the appropriate reviewers, like the legal department. Once the use-case has been approved, your data scientists will be notified and they can begin the process of creating new models in Amazon SageMaker and publish model information to IBM watsonx.governance (figure 4).
Figure 4. Publish Amazon SageMaker model information to IBM watsonx.governance.
Data scientist can continuously prepare data, build, train, test and evaluate the new model until the required accuracy levels are met and a model candidate has been identified. They can register the model using Amazon SageMaker Model Registry and document it using Amazon SageMaker Model Card. This information will be automatically synchronized with IBM watsonx.governance.
The model candidate can be submitted for review and approval in IBM watsonx.governance, prior to deployment (figure 5).
Figure 5. Submit model candidates for approval, prior to deployment.
After the model is approved for production deployment, ML engineer performs necessary configuration and setup for A/B testing and monitoring of the model in production. During the A/B testing cycle, model is monitored across dimensions such as data quality, model quality, bias and explainability are surfaced in the model dashboard and validated against business KPIs. Upon successful validation of KPIs, the model is rolled out for users in production.
Once models have been deployed and optionally associated to an AI Use Case, users can review the status of their associated AI models.
The integration between IBM watsonx.governance and Amazon SageMaker provides customers a unified solution that manages how, where, and when AI is exercised across the organization while maintaining regulatory and compliance obligations.
Conclusion
The integration of IBM watsonx.governance with Amazon SageMaker has been designed to help customers in their journey towards responsible AI. By combining the strengths of AWS’s data AI/ML services and IBM’s expertise in AI governance, this collaboration provides organizations with the tools they need to better meet risk management and compliance requirements, while proactively mitigating risks associated with the growing and changing AI regulations and industry standards.
Whether you’re an existing Amazon SageMaker customer or new to the platform, this integrated offering helps you develop, deploy, and monitor AI models with confidence, enabling conformance to regulations, internal policies, and standard practices every step of the way. Embrace the next era of AI governance and future-proof your business.
Call to action
Contact your AWS or IBM representative today to learn how Amazon SageMaker and IBM watsonx.governance can help you unlock the full potential of your AI initiatives while maintaining the highest standards of risk management and compliance.
Visit the AWS Marketplace for IBM watsonx and other Data and AI solutions on AWS:
- IBM watsonx.data as a Service on AWS
- IBM watsonx.ai
- IBM watsonx Assistant
- IBM watsonx Orchestrate
- IBM Db2 Warehouse as a Service
- IBM Netezza Performance Server as a Service
Additional Content: