It's been nearly a decade since Amazon Web Services (AWS), Amazon's cloud computing division, introduced SageMaker, a platform for creating, training, and deploying AI models. In previous years, AWS has focused on significantly expanding the capabilities of SageMaker, but this year the goal was to streamline it.
At the re:Invent 2024 conference, AWS announced SageMaker Unified Studio, a single place to search and work with data from across your organization. SageMaker Unified Studio integrates tools from other AWS services, including existing SageMaker Studio, to enable customers to discover, prepare, and process data to build models.
“Customers are using data in an increasingly interconnected way, and we're seeing the convergence of analytics and AI,” Swami Sivasubramanian, vice president of data and AI at AWS, said in a statement. Ta. “The next generation of SageMaker brings together the ability to provide customers with all the tools they need for data processing, machine learning model development and training, and generative AI directly within SageMaker.”
SageMaker Unified Studio allows customers to publish and share data, models, apps, and other artifacts with their teams or members of the broader organization. The service exposes data security controls and adjustable permissions, as well as integration with AWS's Bedrock model development platform.
AI is built into SageMaker Unified Studio, specifically Q Developer, Amazon's coding chatbot. In SageMaker Unified Studio, Q Developer can answer questions such as “What data should I use to better understand my product's sales?” or “Generate SQL to calculate total revenue by product category.”
I talked about AWS in my blog post “Q Developer”. [can] SageMaker Unified Studio supports development tasks such as data discovery, coding, SQL generation, and data integration.
In addition to SageMaker Unified Studio, AWS has launched two smaller additions to the SageMaker product family: SageMaker Catalog and SageMaker Lakehouse.
The SageMaker Catalog allows administrators to define and implement access policies for SageMaker AI apps, models, tools, and data using a single permission model with fine-grained controls. SageMaker Lakehouse, on the other hand, provides connectivity from SageMaker and other tools to data stored in AWS data lakes, data warehouses, and enterprise apps.
According to AWS, SageMaker Lakehouse works with any tool that is compatible with the Apache Iceberg standard. Apache Iceberg is an open source format for large-scale analytical tables. Administrators can optionally apply access controls across data across all analytics and AI tools handled by SageMaker Lakehouse.
In a somewhat related development, new integrations allow SageMaker to work better with Software-as-a-Service applications. SageMaker customers can access data from apps like Zendesk and SAP without having to extract, transform, or load the data first.
“Customers may have data spread across multiple data lakes or data warehouses and would benefit from an easy way to integrate all this data,” AWS wrote. “Today, customers can use their favorite analytics and machine learning tools on their data to perform SQL analysis, ad hoc queries, data science, and machine learning, regardless of how and where the data is physically stored. , which can now support use cases such as generative AI.”