Carbon Platform

Carbon Platform is a work project focused on personal data infrastructure for AI systems.

The platform connects to user-authorized data sources, normalizes private context, indexes it, and exposes it through permissioned developer interfaces.

My role

I worked across backend architecture, cloud infrastructure, data modeling, provider integrations, and developer-facing APIs.

The system involved a lot of unglamorous but important problems:

• OAuth flows
• readonly provider scopes
• user consent
• data normalization
• search and retrieval
• event modeling
• infrastructure as code
• security review requirements

Technical shape

The platform used AWS infrastructure with services like Lambda, DynamoDB, Kinesis, S3, OpenSearch, Cognito, API Gateway, CloudFront, WAF, and KMS.

The core data model was built around normalized atomic events: small structured records representing actions, actors, objects, timestamps, provenance, content, and privacy boundaries.

Why it mattered

AI tools are only as useful as the context they can safely access.

Carbon explored what it takes to make personal context available to developers without treating private user data like a loose pile of text blobs.

Notes

This is a work project, so this page intentionally avoids private implementation details, internal code, and sensitive product information.