Skip to Content
Coding House
Talk to Us

AI Learning Platform

Study decks from any source at 75% lower AI cost

Client
Beneficial
Industry
EdTech
Platform
iOS & Android
Timeframe
3 months
AI learning app home screen with generated study decks

About project

How we cut per user AI compute cost by more than 75% and shipped a production EdTech app on iOS and Android in 3 months.

A solo founder had a clear vision and a design concept but no working product. The market gap was real: no mobile-first tool let anyone convert any content source into structured, interactive study material.

Users generate study decks from text prompts, PDFs, images and YouTube videos. The challenge was never the idea. It was making AI generation cheap enough for a competitive subscription price.

App Store

Challenges & Solutions

Challenges

Running AI generation pipelines the obvious way would have made per user compute cost so high that the subscription price could not compete and the business model would not hold. Most partners would have built the naive version and left the cost problem for later. Later would have been too late.

Solutions

We designed the whole system around cost predictability. Cloudflare Workflows orchestrate the AI pipeline, Supabase handles high speed data operations and Firebase covers auth and notifications, so compute costs are fixed and foreseeable.

An optimisation layer routes each task between high capability and lightweight LLMs based on its complexity, and a custom async flow keeps users engaged while generation runs and brings them back with a push notification when their deck is ready.

Main functionality

Building an AI Learning Platform on a Founder's Budget

Multi-source generation
Study decks from prompts, PDFs, photos and YouTube videos.
Async generation flow
Users keep learning while decks generate, then get notified.
Smart model routing
Hard tasks go to large models, simple ones to light models.
University discovery
Browse courses by institution, ready for a B2B channel.
App Store approval for AI
Compliance framing for AI generated content workloads.

Results

75%+
lower AI compute cost per user
3-4x
cost per run avoided
3 mo
from concept to production
  • Full production launch on iOS and Android with App Store approval for AI workloads
  • Per user compute cost more than 75% below the original plan
  • Subscription price kept competitive thanks to the infrastructure savings
  • Platform positioned to scale into universities alongside B2C learners

Technical implementation

  • Serverless, distributed compute layer with fixed and foreseeable costs
  • Cloudflare Workflows for retryable AI pipeline orchestration
  • Supabase for high speed data operations, Firebase for auth and push

Industries

Services

Solutions

Technologies

More use cases

Talk to Us

Have a similar project in mind?

Let's discuss your idea and find the right solution for your business.

Talk to Us