When I set out to build Kaya ML — a no-code machine learning curriculum for UK secondary schools — I gave myself two constraints:
- It had to be built in 5 days
- It had to be built for free
No funding. No team. Just purpose, urgency, and a clear outcome: make real-world AI learning accessible, fun, and classroom-ready.
Here’s how I architected the whole thing — using only free tools.
🔧 The Tech Stack (All Free)
Layer | Tool | Purpose |
Frontend | Notion + Super.so | Website, landing page, CMS |
Design | Figma + Canva | Logo, brand assets, page mockups |
Curriculum Engine | Notion + Google Docs | Writing, planning, resource structure |
Project Layer | Teachable Machine, Google Earth, Blob Opera | Interactive no-code AI tools |
Automation | Google Forms + Sheets | Lead capture, onboarding, waitlist |
Optimisation | Custom CSS in Super | Mobile layout fixes and visual polish |
Chat Layer | Notion AI + third-party embed tools | Support, interactivity, student engagement |
All of these were used at the free tier — with smart constraints, clean logic, and zero tech debt.
🧠 Design Principles
These weren’t just random tools thrown together. I followed a few key principles to make it scalable and solid:
1. Every Tool Has a Job
No overlaps. One tool for each function. Notion for content. Forms for capture. Figma for layout. This kept things modular and easy to update.
2. No-Code, By Design
Kaya ML is for teachers — so the build needed to match the product’s spirit. Everything had to be usable, editable, and non-technical.
3. The Product Is the Content
There’s no backend logic or user data to manage. The lessons are the experience. That made the system lightweight and launchable.
4. Update in Real Time
Any change in Notion instantly updates the live site. No deployment. No dev cycles. Just create, test, ship.
🚀 Why It Works
- No friction — Teachers get a clean interface with no login required (unless needed)
- No maintenance — No database, no servers, no codebase to debug
- High trust — It feels human and crafted, not corporate or bloated
- Scalable — Can layer paid tools later (Stripe, Airtable, analytics, CRM)
💬 Reflection
"Architecture isn’t about code — it’s about thinking clearly under constraints. Kaya ML taught me that good systems come from limits, not from freedom."
This wasn’t a hackathon build. It was a carefully planned ecosystem — designed for delivery, trust, and actual classrooms. And it’s held together by clarity, not capital.