Cloud Shape AI

Inspiration

Modern cloud architectures on AWS can quickly become complex, over-provisioned, and costly. Developers often spend hours analyzing cost reports, configuring auto-scaling, and tweaking resource parameters—time that should be spent building core product features. Cloud Shape AI was created to serve as an intelligent co-pilot for cloud developers, continuously analyzing infrastructure patterns and automatically "shaping" AWS resources for peak performance and minimal cost before deployment.

What it does

Cloud Shape AI is an automated cloud optimization platform powered by Amazon Bedrock and AWS serverless architecture. It integrates directly into developer workflows to:

• Analyze Infrastructure as Code (IaC): Scans Terraform and AWS CloudFormation templates prior to deployment to detect over-provisioned instances, inefficient database tiers, and redundant data paths.
• Intelligent Shape Optimization: Recommends specific, actionable AWS resource adjustments (such as dynamic Graviton instance switching, S3 lifecycle policy rules, and Lambda memory tuning).
• Predictive Cost Modeling: Delivers immediate cost-impact previews before pull requests are merged into production.
• Automated Patch Generation: Generates optimized IaC code patches that developers can apply with a single click.

How we built it

• AI Engine: Leveraged Amazon Bedrock (using Anthropic Claude models) to evaluate cloud topology blueprints and synthesize actionable optimization strategies.
• Backend Services: Built on AWS Lambda and Amazon ECS using Python and TypeScript for serverless, low-latency execution.
• Data & Pricing Telemetry: Utilized Amazon DynamoDB for fast configuration metadata storage and integrated directly with the AWS Cost Explorer API.
• Developer Interface: Developed a clean web dashboard alongside a GitHub Actions plugin for seamless CI/CD pipeline integration.

Challenges we ran into

• Complex IaC Parsing: Mapping diverse Infrastructure-as-Code syntaxes into a standardized format for AI evaluation was challenging. We resolved this by building an intermediate Abstract Syntax Tree (AST) normalizer.
• Accurate Cost Telemetry: AWS pricing dynamically varies across regions and deployment types. We designed an in-memory caching layer to handle high-frequency pricing queries efficiently without hitting rate limits.

Accomplishments that we're proud of

• Achieved sub-3 second analysis times for standard AWS CloudFormation and Terraform stacks.
• Simulated real-world infrastructure benchmarks demonstrating an average 25% to 40% reduction in monthly cloud expenditure.
• Successfully deployed a functional end-to-end prototype integrated directly into GitHub pull request checks.

What we learned

• Best practices for prompt engineering and structured JSON output parsing using Amazon Bedrock.
• Deep architectural nuances in AWS cost optimization, compute family efficiency, and serverless scaling mechanics.

What's next for Cloud Shape AI

• Adding direct read-only AWS IAM integration to scan live, existing cloud environments.
• Implementing autonomous self-healing infrastructure policies triggered via AWS EventBridge.
• Supporting multi-account enterprise governance and custom compliance rule engines.

Built With

• amazon-bedrock
• amazon-dynamodb
• amazon-ecs
• aws-cloudformation
• aws-cost-explorer
• aws-lambda
• github-actions
• python
• terraform
• typescript