XRAYAI

XRAYAI scans a small business's website and writes up what it finds in plain English: a health score, how visible the business is to AI assistants, and what the problems are costing in revenue. It's built for Australian small businesses.

I was the lead contributor on the team — 700+ commits, the top committer — across the Node.js/Express API, the Next.js front end and the PostgreSQL data layer.

Highlights

• Multi-stage scan pipeline orchestrating 20+ AI prompts across five LLM providers (OpenAI, Anthropic Claude, Perplexity, Gemini, xAI Grok)
• Parallelized sequential AI stages into concurrent waves, cutting ~43 seconds off every paid scan
• Multi-provider consensus engine plus a quality gate that detects hallucinated output and regenerates the report
• Stripe one-off reports and monthly subscriptions with signature-verified webhooks, idempotent fulfillment and a rescan-or-refund flow for failed scans
• Ran on Google Cloud (Compute Engine, Cloud SQL, Cloud Storage, Secret Manager) with GitHub Actions CI/CD across dev and production
• Refactored a 1,500-line orchestrator behind a golden-master regression harness; added API rate limiting and removed committed secrets