Skip to content
Services

Six lines of production work

We take on problems where off-the-shelf products hit a ceiling and a thin AI wrapper does not work. We do not build brochure sites, landing pages or routine CRM customisation.

01

AI systems and LLM assistants

Not a wrapper around a chatbot, but working AI systems: RAG over corporate data with source citations, local LLMs for compliance-sensitive work, assistants inside messengers and embedded in CRM. For sales teams, lawyers and support.

Qdrant vLLM Mistral Qwen bge embeddings
Why parsing is the real work
A good fit when
  • → A complex product or a large knowledge base
  • → Off-the-shelf AI add-ons did not stick with your staff
  • → Compliance: the data must not leave your perimeter
  • → It has to integrate with the CRM you already run
02

Custom B2B platforms and portals

Complex B2B systems with multi-role models, custom workflows and integrations. We build what off-the-shelf CRMs cannot hold, where customisation turns into a permanent tax.

FastAPI PostgreSQL React WebSocket Celery
Build or buy: the 30% rule
A good fit when
  • → A partner network that needs its own portal
  • → A workflow that does not fit a ready-made CRM
  • → Several roles with different rights and different interfaces
  • → A regulated industry — legaltech, medtech, fintech
03

Replacing SAP, Oracle and Salesforce

A full replacement of a foreign system — with data migration, processes preserved and on-premise deployment. Run process by process through a parallel track rather than as a single cutover.

FastAPI Accounting integrations RabbitMQ On-premise PostgreSQL
A good fit when
  • → Regulatory or government requirements apply
  • → The vendor left your market and you need a replacement
  • → Integration with your accounting system is the key requirement
  • → On-premise deployment is mandatory
04

Marketplace aggregators and e-commerce systems

Two-way synchronisation of stock and prices across platforms, automated repricing driven by margin rules, unit economics by SKU and channel, alerts on sales anomalies.

FastAPI Marketplace APIs ClickHouse Bots
A good fit when
  • → The brand sells on three or more marketplaces
  • → A catalogue of 200 SKUs or more
  • → Oversells and penalties are a constant problem
  • → You need one view of profitability
05

SaaS platforms and multi-tenant systems

Complete SaaS products with subscription billing, recurring payments, multi-tenant architecture and white-label capability. For companies that have outgrown a stack of no-code tools.

FastAPI PostgreSQL Billing PWA React
Multi-tenant: now or later
A good fit when
  • → The business has hit the ceiling of a no-code setup
  • → A subscription model with recurring payments
  • → Several brands, schools or branches — so, multi-tenant
  • → White-label for other companies is on the horizon
06

Custom analytics and BI

Not a BI tool with stock connectors, but analytics built around your processes. Real-time dashboards for management, demand forecasting with ML, unit economics by SKU, channel or cohort, exports for external reporting.

ClickHouse PostgreSQL Grafana ML pipelines
Measuring a channel nobody reports on
A good fit when
  • → Standard BI tools do not cover your logic
  • → You need real-time analytics, not yesterday's reports
  • → Large data — tens of millions of rows, real load
  • → ML models inside production analytics
Before you commit

What each of these costs, in writing

Rather than a page per keyword, we publish the numbers and the failure modes. Every piece below answers a question clients ask before signing anything.

AI and LLM

AI for triaging inbound requests and email: what to hand the model and where a human is needed

How a language model triages inbound email and requests: topic and urgency, extracting company details and order lines, creating a record in the CRM and routing it to the right person. Where the model gets it wrong and how to measure quality.

Marketplaces

Reconciling the Wildberries sales report and finding money the platform owes you

How to reconcile the weekly Wildberries sales report line by line: sales and returns, logistics, storage, penalties and withholdings, compensations. The errors that come up most often, how to find them through the API and what to do with what you find.

System replacement

Replacing Power BI with a local BI system: what carries over and what gets rebuilt

How to move off Power BI to a Russian BI system or an open-source one: what transfers automatically, why DAX measures and Power Query transformations get rebuilt, why calculations belong in a warehouse, and the order to convert reports in.

B2B platforms

Technical debt: how a client can tell a system is accumulating problems, and what it costs

What technical debt means in a custom system in the client's own terms: the signs you can see without reading code, where it comes from, how much time to set aside for working it off, and when a rewrite is justified.

B2B platforms

The vendor has disappeared and the system is still running: how to take it over

What to do when the developer of your system stops answering or leaves: check access to the server, the domain and the code, take a full backup, find a new vendor and hand the system over. The order of the steps and where the time goes.

AI and LLM

Why nobody uses the AI assistant after launch, and what to do about it

The assistant went live and a month later two people use it: the usual reasons (wrong answers in the first days, an extra step in the workflow, no visible sources) and what to do before and after launch. How to measure whether it is really being used.

FAQ

Before you write

What work do you turn down?

Brochure sites, landing pages and simple customisation of an off-the-shelf CRM. We take projects with complex logic, integrations, AI, data or compliance in them.

Can we start with an audit rather than development?

Yes. For complex projects we usually start with discovery: processes, risks and architecture first, and only then a fixed development scope.

Do you work on-premise?

Yes, and for sensitive data it is often the only option. We deploy on-premise, in your cloud, or on your own servers.

Not seeing your kind of work?

These six are the main ones, but we take adjacent problems where we have the expertise. Describe the problem and we will say honestly whether it is ours.

Discuss a project