Short answer
RaftLabs built online course software for a Bulgarian beauty education startup. Instructors publish courses, learners pay by subscription, and instructors receive payouts through Stripe Connect. Video runs on Mux with DRM, and ElevenLabs handles audio transcription and translation. The platform was built with AI orchestration on a fixed price from February 2026. It's ready for launch but hasn't launched.
Engagement
The engagement
- Client
- A beauty education startup based in Bulgaria
- Sector
- Beauty and salon education
- Decision-makers
- Non-technical founders with a global product vision
- Timeline
- Build from February 2026; ready for launch, not yet launched
- What shipped
- Four connected apps: course publishing, learner subscriptions, DRM video, audio translation, and instructor payouts
- Engagement
- Fixed-price project, built with AI orchestration
- Team
- Developers, QA, and a project manager
- Market
- Bulgaria, with learners and instructors worldwide
- Service
- eLearning platform development
The situation
Beauty and salon professionals learn from people who've done the work. The founders wanted one platform where those educators publish courses, learners anywhere pay to take them, and educators earn from every sale.
That brief breaks a simple course site in three places. The video is the product, so it has to resist screen recording. Learners speak different languages, so audio needs transcription and translation. And money moves two ways: learners pay the platform, and the platform pays instructors.
What makes a paid course marketplace hard
A recorded course is a lost course
On a course marketplace, the video is what learners pay for. If a learner can record it and share it, the instructor's income leaks away, and so does the platform's.
Videos stream through Mux with DRM. DRM playback stops screen recording on supported browsers and devices. Protection lives in a managed video service, not in a custom player someone has to maintain.
One recording, learners in many languages
An instructor records in one language. A learner in another country still needs to follow along. Without translation, every course is limited to one market.
Audio is transcribed and translated with ElevenLabs. For regional audio tracks and captions, we proposed a different design from the original plan. It was easier for learners to use and simpler to build.
Money flows in from learners and out to instructors
Learners pay by subscription. Instructors earn a share of what learners pay. That's two payment problems in one product, and the second one depends on where each instructor lives.
Learner subscriptions run on Stripe, and instructor payouts run on Stripe Connect. Stripe Connect doesn't support every country, which matters for a platform with a global vision. Instructors in an unsupported country can't receive payouts in the first release.
The call
The first release holds to the agreed scope, even where AI made extras cheap
We built the platform with AI orchestration. AI makes features cheap to generate, but every feature is one more flow to test across upload, playback, subscription, and payout. So the first release ships the must-haves, and nothing else.
That meant saying no to work the founders wanted. A revenue split set per instructor stayed out; the platform uses one fixed split. Payout providers for countries Stripe Connect doesn't cover stayed out. So did importing videos from social media, a course review system, and admin screens nobody had asked for. Each is cheap to build and expensive to keep working.
What shipped
The operating loop
One loop: an instructor publishes, a learner pays and watches, and the money finds its way back to the instructor.
The instructor publishes. The learner watches, but can't record.
An instructor uploads course video and publishes it. The platform encodes it, adds translated audio, and serves it with DRM. A subscribed learner watches in their own language. On supported browsers and devices, the player refuses screen capture.
The learner pays the platform. The platform pays the instructor.
A learner's subscription is collected through Stripe. The instructor's share goes out through Stripe Connect at one fixed split between the platform and the instructor. The split isn't something an admin can change per instructor in this release.
Where this applies
Use this when your course is the product and your instructors need paying
This story fits a founder building a paid course marketplace: instructors who earn, learners who subscribe, and video worth protecting. It's the wrong reference if you sell your own courses and a hosted course platform already handles payments and video for you.
A comparable eLearning platform development engagement should start by fixing the payout countries and the feature list before any code is written.
Common questions
Decide where your instructors live, because that picks your payout provider. Decide the revenue split and whether it can vary per instructor. Write a strict in-scope and out-of-scope feature list. And settle the tech stack before development begins. Each of these is cheap to decide early and expensive to change mid-build.
It makes writing code faster. It doesn't make testing faster. AI can produce features that aren't in scope, and each one adds testing work across connected flows like playback and payouts. Hold the build to an agreed feature list and budget for quality work, not just build speed.
This page doesn't state a price. The LMS development cost guide explains how scope, video, and payments change the number before a fixed price is set.
Related work
Related education platform work
TuneClub creator-led music learning app
A creator-led music learning platform for Irish traditional music, built over two years across four releases. 48,100 tracked practice minutes, 100+ learners, and a subscription model built on RevenueCat rather than raw store APIs.
Sekou school management platform and LMS
A French-first school management platform and LMS for K-12 schools in West Africa, built by a six-person RaftLabs team over about six to seven months.
Makeover AI before and after preview tool
Makeover lets service businesses show clients a photorealistic preview of their result on their own photo. Dental, aesthetics, hair, and 40+ other categories use it to convert more undecided consultations.