Amazon SageMaker Book Building Support Services

While many companies are looking to implement AI, it is not easy to establish a real-world operational level AI system within a company. In particular, combining multiple AWS services (Amazon S3, AWS Glue, Amazon Redshift, Amazon Athena, Amazon Bedrock, Amazon DataZone, etc.) to build and operate an integrated AI and data infrastructure requires a high level of expertise On the one hand, Amazon SageMaker is a leading provider of AI and data infrastructure. On the other hand, Amazon SageMaker has evolved into an integrated platform that can natively integrate these services and has the potential to significantly reduce the complexity of building, implementing, and operating it. However, to take full advantage of SageMaker's advanced feature set, systematic knowledge and practical skills in design, construction, and operational design are essential.

As companies seek to fully implement AI, they should address the following issues. Designing and implementing a production configuration that optimally integrates SageMaker and AWS services. 　It is necessary to build an architecture that enables stable operation by linking multiple services such as Glue, Redshift, Athena, Bedrock, S3, etc. in a secure and scalable manner. 　It is necessary to build a stable and operational architecture. MLOps infrastructure including relearning, automatic deployment, and CI/CD must be in place. 　In a production environment, continuous improvement and operation is a prerequisite, not just transient model utilization, so automated re-learning and monitoring pipelines are essential. Designing production operations to meet security and data governance requirements. 　Control designs that meet organizational standards, such as IAM roles, data access control, encryption, audit logs, and metadata management with Amazon DataZone 　IAM roles, data access control, encryption, audit logs, metadata management with Amazon DataZone, etc.

GEAL's “Amazon SageMaker Production Build Support Service” is a professional service to build and operate a data utilization platform for production environments utilizing SageMaker in a secure and sustainable manner. In this service, Zeal will design and implement a highly scalable and controllable architecture that is designed to work with various AWS services (Glue, Redshift, Athena, Bedrock, DataZone, etc.). We will utilize Amazon SageMaker to develop a configuration that enables MLOps, including model development, relearning, automatic deployment, and CI/CD, to run in a production environment, and provide consistent support that goes beyond mere function implementation to operational design, including monitoring, operation, and authority management, as well as skill transfer and knowledge deployment to the operations team and non-engineer layers. We also provide consistent support for the transfer of skills and knowledge to the operations team and non-engineering staff. SageMaker can also be used to incorporate data governance and metadata management mechanisms based on DataZone, accelerating the utilization of data, including AI applications.

Concrete steps Kick-off and assessment of the current situation 　Clarify objectives and use cases 　Inventory of current environment, security, and data flow 　Confirm governance requirements and organizational structure 2. Architecture design and prototyping 　Design integration configuration with Glue, Redshift, S3, Bedrock, etc. using SageMaker 　Study configuration based on non-functional requirements (availability, authorization design, audit support) 　Verify the configuration with a simple prototype (MVP) 3.Implement MLOps and CI/CD configuration 　Develop workflow using SageMaker Pipelines, Projects, Step Functions, etc.