Mappie.ai
An AI SaaS that turns vague product ideas into structured features and specs.

Product teams lose days converting fuzzy ideas into structured features, epics and documentation. The gap between a thought and a spec is mostly manual.
An AI platform that converts ideas into structured specs, with a Notion-like editing experience. As the first engineer on it, I built the core frontend architecture and the AI editing and search experience.
Next.js + TypeScript core architecture
ConvexDB for state and real-time updates
Lexical-based inline AI editor for a Notion-like experience
Embeddings-based vector search over documents
Idea to spec
AI generation of features, epics and documentation from rough input.
Inline AI editor
Notion-like editing built on Lexical with AI assistance.
Vector search
Embeddings-based semantic search across documents.
Document workflows
File upload, parsing pipelines and context-aware AI chat.
| Metric | Value |
|---|---|
| engineer on the product | First |
| of active development | ~6.5 mo |
| live, 1000 spots | Beta |
| in production | Live |
I was the first engineer on Mappie. I built the core frontend architecture, the AI prompt-generation system, the Lexical-based inline AI editor, embeddings-based vector search on ConvexDB, and the document-generation workflows. This was my first end-to-end production AI product.
- Real-time state (ConvexDB) changes how you design collaborative AI UX.
- A great editor is the product - the AI has to feel inline, not bolted on.
