Turnkey product · from idea to production in two regions
UniMe: an operations platform for language schools — one developer, two production regions, 800+ lessons
A school talks to students across several messengers while scheduling, billing and homework live elsewhere. Since February 2026 I have designed and shipped a platform that brings all of it into one window: messengers, CRM, scheduling, homework and an AI assistant. By summer 2026, 3 schools were running on it daily and 800+ lessons had been delivered and billed through it.
At a glance
- Product
- My own, UniMe (unime.tech)
- For
- Language schools and tutors
- Timeline
- Since February 2026, ongoing
- Role
- Architecture, development and operations — solo, implementation through AI
- Languages
- Interface in Russian and English
Results
3 schools
in daily production (July 2026)
158
students and 14 teachers on the platform
800+
lessons delivered and billed through UniMe
35,000+
client messages, about 11,300 a month
26,000+
AI calls in production
2 regions
production in Russia and Europe — Russian users’ data stays in Russia
Before
- Students and parents write in different messengers, so the administrator keeps several apps open at once.
- Leads, scheduling with substitutions and groups, lesson billing and homework live in separate services and spreadsheets.
- A Russian school must keep students’ personal data in Russia, while some of the platform’s users are abroad.
What was built
Messengers in one window
Telegram is connected through TDLib as a full account rather than a bot, plus Max. All of the school’s conversations land in one feed linked to student records.
CRM and leads
A kanban with qualification stages and an AI summary of each contact, so the administrator sees a student’s history without rereading the chat.
Scheduling
Recurring lessons, substitutions, groups, rooms and automatic conflict detection. Delivered lessons go straight into billing.
Homework and knowledge base
Assignments go from issue to review inside the platform. The school’s knowledge base has vector search and an AI assistant that answers from it.
Two production regions
Separate servers in Russia and Europe with a login that routes each user to their region, so Russian students’ data stays in Russia under Federal Law 152-FZ. Build-once deploys with rollback and an automated check of both cells after every release.
Stack
- TypeScript
- Fastify
- PostgreSQL + Drizzle
- pgvector
- Redis + BullMQ
- MinIO
- React + Vite
- TDLib
- Docker, nginx
- OpenAI
Questions
Did one person really build this platform?
Yes. I own the architecture, the task framing and the decisions about what to build; AI writes the code under my control. Since February 2026 the project has 3,000+ commits, all of them mine.
Is Telegram connected through a bot?
No, through TDLib, a full user client. The school works from its real account, and students do not need to install anything or subscribe to a bot.
Why two servers?
Under Russian Federal Law 152-FZ, Russian users’ personal data has to be stored in Russia, and the platform also has users abroad. So production is split into two cells, RF and EU, and the login routes each user to their region.
How many schools use UniMe?
As of July 2026: 3 schools in daily production, with 158 students, 14 teachers, 800+ delivered lessons and 35,000+ client messages.
Can I commission a similar system for my business?
Yes: building web and mobile products from framing to production is one of the three things I do for clients. We start by mapping your processes, not by writing code.