How SAS Viya ensures successful, sustainable model deployments

The speed at which your organization can deploy new AI models determines your ability to stay ahead of the competition.

It’s a measure of your ability to experiment and innovate, try new things and learn quickly. It’s also one of the reasons organizations are investing in SAS® Viya® – which helps you build, manage and deploy models at scale to ensure reliable AI and confident, data-driven decisions.

By accelerating your ability to deploy new AI models, Viya gives you an edge over the competition in an increasingly AI-centric world.

But how does Viya enable this? And what can customers seeking to scale AI-assisted and automated decisioning across their business expect when they invest in Viya? Let’s take a closer look at three key ways Viya makes all the difference in your AI model deployments.

1. Automation and standardization

Viya makes automating and standardizing model deployment processes effortless. The platform offers robust tools for streamlining the development and deployment of AI models, ensuring that they are accurate, efficient and reliable. For example:

• SAS® Model Manager centralizes the management of models built in SAS, Python, R, or other languages. You can register, validate, compare, deploy and monitor models from a single interface. This ensures consistency and governance across the model life cycle to enable faster deployment and reduced risk.
• ModelOps pipelines support continuous integration and delivery (CI/CD) for models, boosting productivity. According to a recent Futurum study, MLOps engineers using SAS Viya were found to be five times more productive than those using alternative platforms.

2. Model monitoring, continuous improvement and adaptability

With Viya, you can continuously improve models with tools that help you regularly revise, retrain or even replace them. This ensures that your models remain accurate and relevant as new data becomes available.

Even better, with Viya’s highly scalable infrastructure, you can retrain models on large datasets more efficiently. This scalability ensures that your models can handle increasing volumes of data without compromising performance. By integrating seamlessly with your existing workflows, Viya makes it easy to incorporate model retraining into your regular processes.

Performance monitoring capabilities in Viya allow you to track the accuracy and efficiency of your models over time. Models can be retrained or replaced automatically using CI/CD pipelines. These pipelines can detect when a model’s performance degrades and trigger retraining using fresh data. You can then deploy your updated models with minimal manual intervention.

Additionally, Viya supports adaptive learning techniques, enabling models to continuously learn and improve as they consume and process new data.

3. Model governance

By incorporating robust governance mechanisms, Viya enhances model reliability and fosters trust and accountability in AI-driven decisions. For example, built-in AI governance in Viya ensures the results of AI models are explainable, transparent and adaptable. Because models are centrally managed, you benefit from consistent model monitoring and traceability, ensuring that the decisions made by AI models are reliable and transparent over time.

Equally important, the platform’s governance capabilities also support ethical AI practices, ensuring that your models operate within defined ethical boundaries and uphold fair decision-making processes.

To help with all of this, Viya includes model cards that act like “nutrition labels” for AI models.

These cards summarize:

• Fairness indicators.
• The model’s purpose and intended use.
• Training data and methodology.
• Performance metrics.
• Governance history and version control.

This makes it easier for you to understand how a model works, why it was chosen and whether it meets ethical and regulatory standards. It’s especially useful in regulated industries like finance and healthcare.

Real stories, real results

In our conversations with customers working in a wide range of industries, it’s clear how Viya is adding value to real-world model deployment processes.

Consider, for example, Georgia-Pacific, a global pulp and paper company. They used Viya to improve efficiency and reduce downtime, especially during the supply chain disruptions caused by the COVID-19 pandemic. Leveraging the results of nearly a terabyte of data flowing through thousands of ML models deployed on and centrally managed by Viya, they were able to make faster, better decisions. These decisions ultimately helped the company improve equipment efficiency and significantly reduce unplanned downtime.