Minimis - AI For Everyone

The Problem

AI literacy is becoming essential — but most UK public schools lack the hardware, software, and funding to provide meaningful access to AI tools. Cloud-based platforms are too expensive or compute-intensive, leaving students underprepared and underexposed.

The Goal

Build a simple, accessible platform where students and teachers can:

• Download and run Small Language Models (SLMs) locally

• Learn and interact with AI without cloud dependencies

• Share creations with a broader learning community

All without needing expensive cloud subscriptions or high-end machines.

The Process

We challenged the default timeline set by the module, and instead ran 4 design sprints in 5 weeks, producing Notion and Framer-based prototypes tested with teachers. In addition to this, I introduced:

• MoSoW-based scoping

• Agile rituals (weekly standups, retros)

• Deliver tracking via Asana & Notion

This helped us prioritize learning fast, scoping correctly, and shipping what mattered.

The Prototypes

First Prototype - made with Notion

A simple landing page so that children can identify what each model does.

Second Prototype - made with Framer

A landing page with a hero and cards for each model.

Third Prototype - made with Framer

This version featured learning kits, which were supposed to help teachers with creating lessons based around a model. We abandoned this version as it went beyond the scope of what the project was about.

Fourth Prototype - made with Framer

Last but not least, this version served as the final source of inspiration for the MVP - pre-populated with sample data of the models as a teaser for what to expect.

The Solution

A full-stack web platform where:

• Teacher and students can upload and download precompiled AI models

• The UI is playful, responsive, and works on low-end devices

• A built-in catalogue, creation system, and moderation tools support classroom use and content quality

Tech Stack

• Frontend: Next.js, TypeScript, Framer, ShadcnUI, Shadcnblocks

• Backend: PostgreSQL, Prisma, Azure Blob, Magic Link Auth

• Infra & Tools: Asana, Notion, Slack, Figma, Whimsical, GitHub

Key Results

• Shipped 93.75% of all scoped features (Must & Should haves)

• All Could-have features delivered as stretch goals

• Demoed live to industry partners and UCL faculty

• Teachers confirmed it supported classroom use and student understanding

• Platform is responsive on low-spac laptops and tablets

Learnings

This project was incredibly challenging, but taught me how the best way to approach ambiguity is through frequent iterations. Along with leading a team for the first time and scoping constraints, I learned to:

• Prototype early (even when the problem space is ill-defined)

• Balance design with accessibility needs

• Scope with discipline and iterate fast based on user and stakeholder input