WAPP
Education

Glinka Digital

How to ship a music edtech startup on the third try: sound quality, failed vendors, and stubborn engineering.

Maksim Ermilov

Maksim Ermilov

WAPP founder · product & architecture

Д

Dima

Full-stack & backend developer

А

Andrey

Mobile Developer

Н

Nastya

Project manager

Ж

Zhenya

QA engineer

Glinka is a long trust cycle: the client came to us three times, left twice for cheaper bids, and came back each time when "faster and cheaper" meant a product that did not work.

How we started: two years and three attempts

For us, the Glinka story began two years ago. We scoped the work and debated architecture; the client still went to other shops. One vendor could not deliver acceptable audio; another could not stabilize the mobile app.

Only on the third pass, when the cost of another failure became too high, we shook hands and went all in. The job was not "write some code" but ship something that would finally fly.

The negotiation and vendor churn phase ran for roughly two years, but the active build on our side took just four months — from fall through spring. That let the client move into testing instead of another endless "almost there" prototype.

The problem: why Zoom breaks music

Mainstream video stacks (Zoom, Skype, Meet) are tuned for speech. Their DSP happily kills anything that looks like "noise": violin partials, piano decay, high vocal notes. For professional teaching that is a hard no.

What we did:

  • Audio without the censorship layer: we dug into WebRTC and turned off the stock noise suppression. The teacher hears the instrument, not a digital caricature.
  • Native modules: off-the-shelf React Native tooling did not give the camera control we needed. We rewrote part of the stack so we could switch to a true wide-angle view — essential to see a pianist's hand technique.

2 yrs

deal cycle

3rd try

the one that shipped

0

audio beauty filters

Stack instead of duct tape

We did not stitch a monster. Matrix handles messaging; a custom WebRTC path carries the lesson. Laravel runs business logic on the server; React Native puts the classroom on an iPad or phone anywhere with a network.

Outcome

The product is live. Right now the first 200 teachers are on a controlled rollout. The blunt lesson: with the right engineering culture, even a "cursed" startup can turn into a business that actually runs.

Product modules

01. Dashboards: one learning ecosystem

We split the product into two focused surfaces. Teachers run groups, read the real schedule, and keep materials under control. Students reach past-session recordings and a sheet-music library in one tap — no scavenger hunt across tools.

A single entry point for how class actually runs: timetable, files, progress, and chat — no admin chrome for its own sake.

Role pick: student or teacher
1 of 3

02. Calls without clutter

This is the core screen. Controls stay out of the way of the picture. One action switches the camera to macro or wide angle so a teacher can correct wrist position or fingering without guessing.

Call UI: HD video and low-latency audio so vocal and instrument lessons feel present — not like a compressed conference line.

Student joining the lesson
1 of 3

03. Built-in metronome & player

Music needs precision. The design system bakes in a tempo-aware metronome and a backing-track player inside the lesson flow. Fewer hand-offs to random apps means fewer sync issues and less “where did that sound come from?”

On-session tools: Metronome and player share the same surface as the call — rhythm, tempo, and rehearsal against a minus track without leaving the session.

Metronome and player in the same lesson surface

Project team

Nastya — Project Manager. She is the person who kept the project with the studio. For two years she steered negotiations, estimate rounds, and client advice while the client tried other vendors. Once engineering started in earnest, she held communication together and kept delivery on schedule.

Andrey — Mobile Developer. He implemented the demanding WebRTC layer on React Native. His main wins were native module work for lens switching and streaming video without noticeable lag.

Dima — Backend Developer. He built the Laravel architecture, integrated Matrix for messaging, and wired S3 so lesson recordings are available without friction.

Evgeniya — QA Engineer. She walked every usage path that mattered: call stability when jumping from Wi‑Fi to 4G, and audio fidelity across tablet and phone models.

Maksim — Founder / CTO. Technical oversight, stack choice (WebRTC + Matrix), and final product quality control.