Discuss your project
Services Products How we work Courses Estimate Contact Discuss your project
Automation engineering

We automate the work your team still does by hand

AI automation, backend services, bots and web platforms. Not slideware demos — systems in production: queues, monitoring, resilience, and code your own team can maintain.

Discuss your project Get an estimate

We reply within one business day. The first call takes 30 minutes and is free.

example: processing a sales call
$ POST /v1/calls ← call recording (mp3)
< { "job_id": "0199c4e2…", "status": "queued" }
· worker: whisperx → transcript + speaker labels
· llm: summary, topics, commitments, next step
✓ webhook → CRM: deal updated, follow-up task created
· nobody fills in the CRM card by hand
2 of our own products running in production
2–4 wks from kickoff to a pilot on real data
100% code and infrastructure stay yours
24/7 monitoring and alerting on support plans
Services

What we do

We take on the whole process: from understanding how it actually works today to a running system and a trained team.

AI process automation

We find the places where people move data around by hand and remove the manual work that creates no value.

  • Speech recognition: calls, meetings and audio turned into text with speaker labels
  • LLM assistants grounded in your knowledge base (RAG), not in the open internet
  • Extracting structured data from documents, emails and inbound requests
  • Classification and routing of requests, automatic summaries and next steps
  • Human in the loop: edge cases go to an operator instead of being silently guessed
LLMRAGSpeech-to-TextpgvectorQdrant

Web development and platforms

Sites and products that load fast, get indexed, and don't fall apart in their second year.

  • Corporate sites and landing pages with server-side rendering and honest SEO
  • Customer portals, dashboards and internal tools
  • Product platforms: catalogues, learning, marketplaces, billing
  • Accessibility, mobile, dark mode, Core Web Vitals in the green
  • Multilingual setups across several domains
GoTypeScriptReactSSRCore Web Vitals

Bots and messaging

An entry point where your customer already is: requests, payments and support inside the messenger.

  • Telegram and MAX: flows, forms, payments, notifications
  • Website requests land in your team channel; a manager takes over with one tap
  • AI answers the routine, a human picks up the hard cases — without losing context
  • Integration with CRM, spreadsheets, calendars and internal services
  • Admin panel: who took the request, what was answered, how it ended
Telegram Bot APIMAXWebhooksCRM

Services, integrations, operations

A backend that holds up under load, and infrastructure where you can see what is going on.

  • APIs and asynchronous services: queues, idempotency, signed webhooks
  • System-to-system integrations and zero-downtime data migrations
  • Deployment: Docker Swarm or Kubernetes, CI/CD, one-click rollback
  • Monitoring, logs, alerting, incident review
  • Support under an SLA — or a clean handover to your team
Docker SwarmCI/CDPrometheusGrafanaLoki
Products

We build for ourselves first

Two products we designed, wrote and operate ourselves. Not a portfolio of screenshots — running systems where we test our own decisions.

Call transcription

Copione YB

in production

A B2B API that turns a call recording into text with speaker labels. Asynchronous, with guaranteed delivery of the result.

API documentation
  • Accepts an uploaded file or a link to an object in your own storage — audio never has to be copied to us
  • Diarisation: who spoke and when, with modes for telephony and for meeting recordings
  • Multi-tenant: separate keys, rate limits and signed webhooks per client
  • Idempotency: re-sending the same recording never creates a second job
  • Audio kept under a retention policy — recordings don't outlive their purpose
GoPythonWhisperXRabbitMQPostgreSQLMinIO
Online learning platform

OnlineSchool

in production

A full LMS with live classes: courses and lessons, graded homework, video, a shared whiteboard and chats.

Open the platform
  • Catalogue of tracks and courses, modules, lessons, materials, progress
  • Homework: submission, teacher review, grades
  • Live classes with video and attendance, schedule and groups
  • Shared whiteboard and class chat — in real time
  • Notifications, analytics and an audit trail
GoReactLiveKitPostgreSQLRedisS3
How we work

Five steps instead of one big promise

Every stage ends with something you can look at and verify. If the pilot produces no effect, you see it in the numbers — not a year later.

Understanding the problem

A 30-minute call. We look at how the process works today and where time and money leak. Sometimes it turns out no software is needed — and we say so directly.

Estimate and plan

We fix the scope, the timeline and the price, and write a specification you can actually accept work against instead of judging by feel.

Pilot

In 2–4 weeks we ship a narrow slice on real data. The goal is a measurable result: hours saved, manual steps removed.

Rollout

We scale it across the process, integrate with your systems, train the team and hand over documentation and access.

Operations

Monitoring, alerting and further development under an SLA. Or a full handover to your team — the code and infrastructure are yours anyway.

We work in stages with a fixed price per stage. You can start with a single pilot, without a long contract.

Technology

What we build on

Boring, proven tools where reliability matters; modern ones where they give a real advantage.

Backend

GoPythonPostgreSQLRedisRabbitMQsqlcRiver

AI and data

LLMRAGWhisperXpyannoteQdrantpgvector

Frontend

TypeScriptReacttemplViteSSR

Infrastructure

Docker SwarmKubernetesTraefikMinIO / S3PrometheusGrafanaLokiCI/CD
Learning

Our own online course platform

We don't only build learning platforms — we teach on ours.

OnlineSchool is our platform: live classes with video and a shared whiteboard, homework reviewed by a teacher, progress tracking and analytics. Courses on development, automation and the practical use of AI.

Open the platform

The course catalogue, enrolment and student area live on the platform.

  • Live classes and recordings
  • Practice with reviewed homework
  • Shared whiteboard and chat with the teacher
  • Progress, grades and attendance
  • Groups and schedule
  • Corporate training for teams
Estimate

A rough number in under a minute

Pick what you need and we'll show a realistic range and timeline. It's a planning figure, not a quote.

What you need

Scope of the first stage

Design

After launch

Intro call

30 minutes to find out whether there's a project here

Pick a time that suits you. The call is free and commits you to nothing.

  • We go through your problem and its context
  • We show how we solved something similar
  • We propose 2–3 options with estimates
  • No decks, no sales pitch
No free slots right now — write to us and we'll arrange a time.

Request received

We've got it and will reply within one business day.

Contact

Tell us what needs automating

A couple of sentences about the problem is enough — we'll reply within one business day with a concrete next step.

Telegram @ybdev_botEmail hello@ybdev.ruOur Git git.ybdev.ru

Request received

We've got it and will reply within one business day.

FAQ

What people usually ask

How much does it cost?

The calculator above gives a range based on comparable projects. The exact price appears after we look at the problem — the first call usually shows whether a smaller scope would already do.

How long does it take to launch?

A pilot on real data takes 2–4 weeks. A full rollout depends on how many integrations are involved, but we deliberately cut the first stage down to the minimum that is already useful.

Who owns the code?

You do. Repositories, infrastructure and documentation are handed over in full. We can deploy the system inside your own perimeter and hold no access to it after delivery.

What about data security?

By default everything runs on your own servers: databases, storage, speech recognition models. If a task genuinely needs an external LLM provider, we raise it explicitly and decide together — nothing leaves your perimeter quietly.

Do you support the system after delivery?

Yes, under an SLA: monitoring, incident response, further development. It's a separate agreement, not a condition — you can take the system and run it yourself.

Do you work with clients outside Russia?

Yes. We run projects in Russian, English and Italian.