SaaS architecture, UI engineering, AI integration
Google Tag Manager, without the tag manager pain
Built TagRabbit, a drag-and-drop and LLM-driven Google Tag Manager configurator.
The problem
Configuring GTM means wiring variables, triggers and tags by hand — error-prone work that marketers depend on engineers for.
What I did
- SvelteKit drag-and-drop editor that maps visual blocks onto a GTM configuration model.
- A translation layer that compiles that model into variables, triggers and tags.
- Stape.io integration for server-side tracking.
- An LLM path (Gemini, OpenAI, OpenRouter) that produces the same model from plain language, orchestrated with Mastra and FastMCP on a Django + RabbitMQ backend.
What changed
- Shipped as a production SaaS with two input modes — visual and natural language.