I built a fully serverless AWS learning platform with Claude — here's how it went

I recently embarked on creating HomeServerLab, a free AWS learning platform designed for developers, utilizing exclusively serverless infrastructure. The technology stack consisted of AWS Lambda powered by Python to handle all routes, eliminating the need for traditional web servers, along with API Gateway HTTP API as the entry point for all requests.

DynamoDB was employed for rate limiting, user management, chat history, and OAuth state, while Amazon Bedrock (Nova Micro) served as the driving force behind the integrated AI assistant. Cloudflare played a pivotal role in managing DNS, CDN, WAF, and bot protection, alongside Cloudflare R2 for apps marketplace storage. Remarkably, the entire backend operates within a single Lambda function, incurring minimal costs—under $1 per month.

Throughout the development process, I relied heavily on Claude. From crafting the JWT authentication flow to troubleshooting DynamoDB rate limiting challenges, debugging OAuth GitHub login issues, and enhancing SEO with structured data and Content Signals for AI crawlers, Claude's assistance proved invaluable. These decisions often lacked clear right or wrong answers, requiring thoughtful judgment.

For instance, determining the placement of CommonRuleSet in Count vs Block on AWS WAF or structuring Cloudflare caching rules without disrupting the OAuth flow. HomeServerLab was built by a single developer based in Portugal.