OnMicro.AI
OnMicro.AI is a user-friendly platform that allows educators to build and deploy AI applications as easily as creating a Google form. The platform's mission is to provide agency over AI for educators through accessible tools and research, putting them in the driver's seat rather than treating AI as a black box of complex algorithms.
US
Overview
When John came to Proga Tech, onMicroAI had real traction but a set of compounding technical problems that were slowing the product down. The issues touched different layers of the platform at once, which made them difficult to address in isolation.
Challenges
01/
Payments infrastructure was absent
The platform had no working payment system in place. There was no clear path for users to subscribe or pay, and no technical documentation to guide an implementation in the right direction.
02/
The core builder was unintuitive
The app builder sits at the heart of the platform, yet it was difficult to use and had accumulated enough UX debt that it was becoming a barrier for the educators it was meant to serve.
03/
No automated test coverage existed
With no end-to-end testing in place, every deployment carried an implicit risk of regression. There was no safety net to catch problems in core functionality before they reached users.
04/
The AI component was costly and the pages needed attention
The AI processing pipeline for file resources was generating significant costs without an optimization strategy, while statistics pages were outdated and overall platform performance had room to improve.
Solutions
Our team approached each problem with the same discipline: understand the issue properly before proposing a solution, align on direction before writing code, and document the reasoning so decisions can be revisited later. The result was a structured engagement that moved through all four workstreams without losing coherence.
[01]
Stripe and payment integration
We began with a technical consultation to map the right approach for John's specific use case, then produced full technical documentation including a step-by-step implementation plan. Once the plan was agreed upon, we carried out the implementation end-to-end, from API configuration through to the checkout flow.
[02]
App builder redesign
Rather than patching over the existing interface, we held structured discussions with John to align on a new UX direction and surface improvement proposals. We prepared revised designs before writing any code, then implemented the redesigned builder and added new features to extend its capabilities beyond the original scope.
[03]
Automated testing
We introduced end-to-end tests with Playwright to cover the platform's core user flows, and configured a continuous integration workflow so the full test suite runs automatically after every deployment. Regressions are now caught before users encounter them.
[04]
AI performance and cost optimization
We researched optimization strategies for the AI component's file-processing pipeline, then implemented a Retrieval-Augmented Generation (RAG) system which reduces the number of tokens used for AI analysis when working with file resources. The pages were redesigned and statistics updated in parallel, giving the platform a more cohesive and reliable presentation throughout.
Results
Each workstream delivered concrete, verifiable improvements. The most significant quantitative result came from the RAG implementation, which reduced AI costs on file resources by several times compared to the previous approach.
The platform now has a fully operational payment infrastructure where none existed before, and every deployment runs against a suite of automated end-to-end tests covering the core functionality. The builder, which was the most critical piece of the product, went from a recognised pain point to a substantially improved experience that John and his users interact with every day.
5+
Reduction in AI cost
100%
Better OnMicro.AI onboarding & app builder functionality
30x
Projected AI cost reduction for large-file processing
Technology stack
The project utilizes cutting-edge technologies to provide structured and reliable solutions to contemporary challenges.
Frontend
Next.js
For routing, SSR/SSG, and performance
TypeScript
For typed, maintainable UI code
Tailwind CSS
For layout and accessible components
Backend
LiteLLM
(And provider SDKs such as OpenAI / Anthropic) for model callse
PostgreSQL
As the primary relational datastore
Python
Django 5 and Django REST Framework for APIs and business logic
Payments
As Stripe Payment Gateway integration experts, we implemented a comprehensive billing system for OnMicro.AI:
- Monthly subscription billing with limited AI tokens usage
- Paid add-ons to bump up tokens after the subscription’s AI tokens limit is over
- Safety guards for AI tokens balance
Build mode
Create your own mini AI-powered app
Preview mode
See it in action in preview or app modes.
In John's words
“
The Proga team have been a fantastic partner for OnMicro! Two things stand out: First, the ability to consistently turn raw vision into well-crafted requirements. Second, the ability to leverage AI for speed but without losing control or accuracy. We’re able to move quickly while minimizing technical debt and maximizing everyone’s understanding of how things work.
Beyond project delivery
Advanced topics require offline brainstorming. Andrew and Mia met in Warsaw to discuss AI prompts branching on edit. Do we make it a tree-based branching, or do we make it a flat side-by-side versioning? The decision was not to overcomplicate things until our users ask for it, so a simpler version made it to production.
"
Keep things simple
Proga team says
Is your platform facing similar challenges?
Proga Tech works with product teams to turn technical debt into momentum, whether the priority is payments, UX, test coverage, or AI cost optimization. Get in touch and tell us what you are working on.
Contact us
