Deploy this Guidance
Use sample code to deploy this Guidance in your AWS account
Overview
This Guidance automates the selection of environmental impact factors (EIFs)—sometimes also referred to as emission factors—using foundation models with retrieval augmented generation (RAG) to scale product carbon footprint assessments. EIFs measure activities into metrics to assess potential environmental impacts, like carbon dioxide equivalent (CO2e). The Guidance takes in activity data descriptions (such as purchase orders) and parameters (such as regions), and returns a ranked list of EIFs with similarity scores. Human annotators, invoked by low similarity scores or random sampling, provide ongoing evaluation, improvement, and corrective annotations.
The Guidance is based on using two algorithms published by Amazon:
CaML
How it works
Option 1
This architecture diagram shows how to scale carbon footprint assessments using a knowledge store made up of Amazon S3 and Amazon Bedrock. For Option 2, open the other tab.
Download the architecture diagram
Step 1
Collect business activity data and emission factors from appropriate sources.
Step 2
Index relevant activity category datasets using knowledge bases for Amazon Bedrock to build vector knowledge stores for searching during mapping.
Step 3
Ingest the activity data to Amazon Simple Storage Service (Amazon S3). AWS Glue cleans and formats incoming data for mapping activities to impact factors.
Step 4
AWS Step Functions iterates through the business activities. It calls large language models (LLMs) hosted by Amazon Bedrock to match the given activity to the most relevant emission factor category using RAG. Matches are stored in Amazon S3.
Step 5
AWS Glue combines the matched activity and category with its corresponding kg CO2e from relevant emission factor datasets. This combined dataset is registered in AWS Glue Data Catalog.
Step 6
Report on mapped emission factor data with Amazon QuickSight or Amazon Athena.
Option 2
This architecture diagram shows how to scale carbon footprint assessments using a knowledge store made up of Amazon OpenSearch Service Neural Search. For Option 1, open the other tab.
Download the architecture diagram
Step 1
Collect business activity data and emission factors from appropriate sources.
Step 2
Use Amazon OpenSearch Service Neural Search feature to convert emission factor query text into vector embeddings, and return close matches.
Step 3
Ingest the activity data to Amazon S3. AWS Glue cleans and formats incoming data for mapping activities to impact factors.
Step 4
Step Functions iterates through the business activities. It calls large language models (LLMs) hosted by Amazon Bedrock to match the given activity to the most relevant emission factor category using RAG. Matches are stored in Amazon S3.
Step 5
AWS Glue combines the matched activity and category with its corresponding kg CO2e from relevant emission factor datasets. This combined dataset is registered in Data Catalog.
Step 6
Report on mapped emission factor data with QuickSight or Athena.
Deploy with confidence
Everything you need to launch this Guidance in your account is right here.
Well-Architected Pillars
The architecture diagram above is an example of a Solution created with Well-Architected best practices in mind. To be fully Well-Architected, you should follow as many Well-Architected best practices as possible.
Operational Excellence
Amazon CloudWatch provides centralized logging with metrics and alarms across all deployed services to raise alerts for operational anomalies.
Security
Resources are protected using AWS Identity and Access Management (IAM) policies and principles. Use least-privilege access and role-based access to grant permissions to operators. Data at rest is encrypted using AWS Key Management Service (KMS). HTTPS endpoints with transport layer security (TLS) provide encryption in transit for service endpoints.
Reliability
These serverless services automatically adapt to demand changes. Amazon S3 is fully elastic, growing and shrinking as data is added or removed, providing durable cloud storage and industry-leading availability without overprovisioning. Step Functions offers serverless orchestration with built-in error handling for modern applications. As the application runs, it maintains state, tracks workflow steps, and stores an event log passed between components. If networks fail or components hang, the application can resume from the last checkpoint, ensuring a seamless workflow.
Performance Efficiency
Amazon Bedrock allows developers to choose suitable foundation models from leading AI companies for their use cases. Amazon Bedrock is a fully managed service offering high-performing foundation models through a single API, along with capabilities for building secure, private, and responsible generative AI applications.
AWS Glue enables quick data discovery and preparation for analytics, ML, and application development. AWS Glue is a serverless data integration service, eliminating infrastructure management through automatic provisioning and supporting various data processing frameworks and workloads.
Cost Optimization
This Guidance relies on serverless services such as Amazon S3, Athena, and AWS Glue. These services require no infrastructure setup or management and scale automatically to match demand, ensuring minimum resource utilization. Serverless scaling helps reduce overall resource usage and costs by adjusting to fluctuating demands.
Sustainability
Step Functions call Amazon Bedrock directly, eliminating the need for dedicated compute to process data. This approach reduces compute usage and, as a result, lowers carbon footprint.