Inspiration
Most startups fail not because the idea is bad, but because validation is slow and iteration is weak. Founders build static landing pages and rarely evolve their positioning based on real signals.
We were inspired to build a system where startup ideas are treated as living entities continuously refined by data, trends, and feedback.
What it does
EvoFounder AI is a self-improving agent that: - Takes structured startup input (name, audience, tone) - Generates a startup concept and mockup landing page - Produces deployable HTML/CSS/JS - Uses trend and performance signals to suggest strategic refinements
Instead of treating ideas as static, it models them as evolving systems.
How we built it
- n8n for workflow orchestration
- AI Agent (Bedrock/OpenAI) for structured idea and code generation
- Custom prompt engineering to enforce structured JSON output
- Dynamic HTML rendering via generated code
- Modular architecture designed to plug into analytics & feedback loops
Challenges we ran into
- Bedrock IAM & Marketplace model access restrictions
- Rate limits and token constraints during testing
- Ensuring the AI returns clean, structured JSON
- Rendering generated HTML dynamically within n8n forms
- Designing around limited UI tools without building a frontend
Thankfully the Ruya CEO was there to help us out with the UI and backend issues :)
Accomplishments that we're proud of
Built a fully working end-to-end AI generation pipeline in one day and successfully generated deployable landing page code from form input
What we learned
Workflow orchestration is as important as model choice
What's next for EvoFounders
ntegrate real analytics (CTR, engagement, bounce rate) and connect to trend APIs and market data sources
Log in or sign up for Devpost to join the conversation.