Streamline your machine learning model deployment lifecycle with Terraform
summary
End-to-end CI/CD automates model provisioning, customization, indexing and guardrails using Terraform modules that run validation, linting, plan and apply in a pipeline tied to git; it provisions Bedrock models, trains a Bedrock custom model from S3 data, builds a vector index on OpenSearch serverless, and wires Bedrock agents and guardrails via the AWS Cloud Control provider because those resources aren’t in the classic AWS provider. The demo provisions into a VPC, completes training and serves queries; integration with Terraform Cloud and ServiceNow enables governed self-service.
AI-written summary · about
chapters (10)
- 0:00Introduction to session
- 1:16ML deployment challenges
- 4:07Key AI and generative AI concepts
- 8:23Amazon Bedrock overview
- 12:46Infrastructure as code with Terraform
- 16:42AWS providers comparison
- 17:05AI banking assistant use case
- 18:09Solution architecture with Terraform
- 21:44Demo walkthrough
- 24:58Summary and resources
related talks
27mFrom Training to Serving: Machine Learning Models with Terraform
01h 30mUnleashing the Power of AWS: A Workshop on Bedrock Assistant
55mSelf-service infrastructure is no longer a dream
19mSynchronization of Infrastructure Provisioning & Application Deployments
10mInfrastructure as a Code Best Practices with Terraform
42mHands-on with the Terraform Module for AWS Landing Zone
57mLook Ma, no OPS! Automating microservices' infrastructure
31mMaking AI work for you: Terraform engineer blueprint