What running an EdTech platform at scale taught us about building one
Lessons from running the engineering behind a one-to-one tutoring platform serving twenty-plus countries: the schedule is the product, the live class must always work, and parents want to see, not ask.
Since 2025 we have run the entire engineering behind Qlive, a one-to-one online tutoring platform for students from kindergarten to Grade 12, in more than twenty countries. Thousands of classes a week go through systems we built and operate. That is a different education from building a learning product on a whiteboard. Here is some of what it taught us.
The schedule is the product
Everyone talks about the classroom. The thing that actually decides whether a tutoring business works is the schedule: matching students with tutors, across time zones, around school hours and exams and holidays, with reschedules and cancellations and the tutor who is unwell at short notice.
If the schedule is right, everything downstream is calm. If it is wrong, every class is a phone call. We now treat scheduling as the core system, not a feature, and we build the exception handling first: what happens when someone does not turn up, when a slot needs moving, when two changes collide.
The live moment has to work every single time
A class at four o’clock is a promise made to a child and a parent. There is no “try again later”. That reshapes engineering priorities: reliability over richness, tested fallbacks, monitoring that tells us before the parent does. Interactive activities and clever features earn their place only once the basic session is boringly dependable.
It also means we build and run, not build and hand over. The team that wrote the classroom is the team watching it at four o’clock.
Parents want to see, not ask
Most of the messages a tutoring business receives are questions a parent could answer themselves if they could see: what is the schedule this week, did the class happen, how is my child doing, what was covered. A parent app that shows progress, schedules and reports without anyone asking removes a large share of the support load and, more importantly, builds trust. Families who can see are families who stay.
We now think of the parent view as a first-class product, not a report bolted on at the end.
Teachers need tools that take work away
A teacher tool that adds forms to fill in after every class is a tool that gets ignored. The ones that work take work away: attendance in one tap, notes captured from what already happened, reports drafted from real activity for the teacher to edit. AI helps here, drafting and summarising, with the teacher deciding what is said about a child. Never the other way round.
Operations is most of the software
The learner sees a fraction of the system. The team behind it lives in the rest: tutor onboarding, quality, payments, attendance, reports, parent communication, the exceptions that make up a real week. On Qlive that is the Connect platform: separate products for operations, teachers, learners and parents, working as one. The operations product got as much design attention as the learner’s, and it shows in how quiet the support channel is.
Twenty countries means twenty kinds of normal
Time zones, obviously, but also school calendars, exam seasons, holidays, payment methods, and what a parent expects a report to look like. Building for one market and “adding” others later is painful. Assuming from the start that every date, time, currency and calendar is local saves a great deal of rework.
AI, in its place
Reading conversations and voice notes so that a promise made in a chat becomes a scheduled action. Drafting for teachers. Answering the routine parent question at eleven at night. Each of these saves real hours. None of them grades a child, and none of them decides anything about a child’s education on its own. That line is written down, and we hold to it.
What we would tell someone building one
Start with the schedule and the operations tooling, even though the classroom is more exciting. Make the live session boring before making it rich. Give parents a window before they ask for one. Build tools for teachers that remove work. Assume every market is different from the first day. And plan to run it, not just to ship it, because an education product is never finished.
If you are building or running one, we are happy to compare notes. The rest of what we do for education companies is on the EdTech software development page.