Case Study
PurpleProof
Serverless backend for an AI-assisted content platform — brand voice, RSS/radar discovery, draft generation, approval flow and multi-platform publishing.
Problem
An AI content platform needed a reliable serverless backend for radar discovery, brand-voice generation, draft approval and multi-platform publishing — including scheduled publish, retries and blockchain sealing — without coupling the API to a fragile monolith.
My role
Internship at AMPD Labs in a team of four. Primary backend ownership: Lambda API design (posts, drafts, brand voice, radar, publishing, auth), DynamoDB data layer, publishing pipeline with OAuth2 and retry/DLQ patterns, JWT roles, and Serverless Framework deploy/ops. Frontend and smart contracts were primarily owned by teammates. Public repo is a sanitized portfolio snapshot — AMPD Labs owns the production IP.
Architecture
Technical decisions
Serverless on AWS Lambda
API Gateway + Lambda with the Serverless Framework fit bursty AI and publishing workloads, kept services independently deployable, and matched the team's need for a typed Node.js/TypeScript backend.
DynamoDB access patterns first
Posts, drafts, brand voices and radar items were modeled around concrete access patterns so draft editing, approval and scheduled publish stayed predictable under Lambda concurrency.
Publishing with retries and DLQ
Multi-platform publish (X/Twitter OAuth, LinkedIn, Discord) used scheduled jobs with retry and dead-letter handling so transient integration failures did not silently drop posts.
Challenges
Integrating AI generation with approval flow
Radar discovery and OpenAI draft generation had to land in editable, role-aware draft states before publish — not fire-and-forget content dumps.
Keeping team boundaries clean
Backend APIs and integration docs had to stay stable for frontend and contract teammates while the product moved quickly across brand voice, radar and sealing hooks.
Outcome
A serverless TypeScript API powering PurpleProof's AI content workflows at AMPD Labs — from radar and brand voice through approval and multi-platform publish, with blockchain sealing hooks. Portfolio materials show a sanitized backend snapshot of that work.