Skip to the content
Cases

Six projects
and what they produced in numbers.

Pilot, client and internal. Each one carries the problem in the client’s own words, what we did, and what we measured the result with. We hold back company names until they give us permission.

Industry
Service
Technology

10 cases

01HoReCaClient project

A Telegram Mini App with pre-ordering and payment

The morning peak queue was running to 12 minutes. We moved ordering and payment into Telegram so the guest collects a cup that is already made.

−73%time spent handling an order
Telegram Mini AppsPythonaiogramPostgreSQL
02LogisticsClient project

Automating how requests are assigned to drivers

The dispatcher was sorting 200 requests a day by hand in a spreadsheet. We replaced route building with a script that checks the constraints.

6 hoursa day freed up for the dispatcher
PythonPostgreSQLRedisn8n
03EducationPilot

An AI agent on first-line support

Before a cohort started, support was taking 900 requests a day. The agent answers from the knowledge base with a link to the source.

68%of requests closed without an operator
ClaudeLangChainQdrantFastAPI
04RetailClient project

A security audit before card acquiring went live

The bank asked for a test report. We found access to other customers’ orders via a direct link, and a promo code that applied twice.

14findings, 3 critical
OWASP Top-10Burp SuiteSemgrepTrivy
05Business softwareClient project

Refactoring 40,000 lines of inherited code

The only developer had left and no documentation remained. We reconstructed a map of the system and made changes safe to ship.

40,000lines worked through in 3 weeks
PythonTypeScriptPostgreSQLDocker
06ResearchInternal project

A rig for testing how well AI agents hold up

We built a lab where our own and our clients’ agents are run through a catalogue of attacks, with a regression suite kept alongside it.

2400+attacks in the catalogue
PythonGarakPyRITClaude
07RetailClient project

An AI sales manager for Avito

A stream of enquiries across several Avito stores, managers unable to reply fast enough, and leads lost to the delay. We handed the first contact to an AI agent.

24/7first contact with no operator
LLMAvitoPythonPostgreSQL
08OperationsClient project

Monitoring and routing of incoming mail

Important letters — price lists, changed terms, urgent requests — got lost in the flow, and staff sorted mailboxes by hand. We set up automatic mail monitoring.

0manual mailbox sorting
PythonIMAPTelegram Bot API
09FinanceClient project

An internal payment accounting and control system

Invoices and payments were tracked in scattered places, and reconciliation with accounting cost hours and bred errors. We built storage, search and automatic reconciliation against 1C.

Autopayment reconciliation with 1C
PythonSQLite1C
10OperationsConcept

Procurement automation and 1C synchronisation

Prices, stock and invoices live in separate systems and need manual merging for every purchasing decision. We worked through the concept of a single loop tied to 1C.

Concepta single procurement loop with 1C
1CPythonвнутренние API

A similar problem?
Tell us what is not working.

After the brief we tell you how long it takes and what it costs. If it is not worth doing, we say that too.