Building a Lead-Generating AI Website Generator

When scaling outbound sales efforts, having hyper-personalized landing pages can dramatically increase conversion rates. However, manually coding and deploying a new page for every target segment is incredibly slow and expensive.

To solve this, I built an automated AI Website Generator that can spin up customized, production-ready landing pages in seconds based on a simple prompt or CRM data trigger.

The Architecture

The pipeline consists of three main components:

1. The Brain (LLMs): We use a fine-tuned LLM to take minimal input (e.g., “SaaS tool for dentists in Chicago”) and generate high-converting copy, structural JSON representing the page layout, and specific Tailwind CSS classes for styling.
2. The Engine (FastAPI): A robust FastAPI backend receives the generated JSON payload. It validates the schema using Pydantic, instantly mapping the data to pre-built Astro components, and injects the dynamic copy.
3. The Orchestrator (n8n): To tie it all together without writing brittle custom glue code, an n8n workflow listens for webhooks from our CRM. When a new high-value lead is added, n8n triggers the FastAPI endpoint, compiles the site, and deploys it straight to the edge.

Real-World Impact

By treating infrastructure as code and leveraging AI for the creative heavy lifting, this system eliminated weeks of manual developer effort. More importantly, it allowed the sales team to experiment with hyper-niche messaging, resulting in a significantly higher meeting booking rate.

Curious about the code? Check out my GitHub for more open-source backend patterns!