AI chatbots on your catalog
A consultant bot for your website or Telegram that answers customer questions from your product catalog and can pass the conversation to a person when it is not sure.
AI chatbots and automations that take routine work off your team
I’m Oleksandr Bratuta, a freelance developer in Milan. I build chatbots that answer from your own catalog and documents, workflows that move data between your tools, and scrapers that collect data on a schedule.
Email me: oleksandr@bratuta.net
What I do
Document assistants with sources
Ask questions about manuals, price lists or internal documents and get an answer with the document and page it came from. If the documents don’t contain the answer, it says so.
n8n and Python automation
Workflows that take in leads, write reports, read invoices and move data between your tools, so nobody has to copy and paste.
Scheduled scrapers and data pipelines
Prices, listings or supplier data collected on a schedule, cleaned and checked, then delivered to a spreadsheet, a database or your inbox, with an alert when something changes.
Demos you can check
These are demo projects, not client work. They run on public or practice data. Every screenshot was rendered from a real run’s output. No numbers were edited by hand.
Demo
Document Q&A assistant
- Four public documents in Italian and English.
- 12 test questions, scored by code: 11 of 12 passed, with the same total in 4 runs.
- On all 4 questions with no answer in the documents, it refused to guess, in the language of the question.
- The one miss: the right page was not retrieved, so it refused instead of inventing an answer.
Demo
Three n8n business workflows
- Lead intake: a website form goes to an AI that sorts it by intent and urgency, then it’s logged and a draft reply is written for a person to check. Spam is marked low priority and invalid input is rejected.
- Daily sales report: KPIs from the orders export (two days of sample data shown side by side) and a separate alert branch for refunds. That branch fired in the run of 12 October 2026.
- Invoice PDFs to spreadsheet: 6 invoices across 2 runs. One was flagged for review by code.
Demo
Scheduled scraper with change alerts
- 1,000 books (products) and 100 quotes collected in one run: 141 requests, 70.3 seconds, 0 invalid rows.
- Output to CSV, XLSX and SQLite, ready for Google Sheets.
- Each run is compared with the previous one: new, removed, price, rating and stock changes.
- The second run found 0 changes because the practice site never changes. The comparison logic is covered by a unit test.
Built and tested
Telegram consultant bot on a real product catalog
A Telegram bot that answers questions about a real building-materials catalog. It is built and tested, but no paying customers use it yet.
How I work
- A short call or email exchange. We look at what is slow or manual today and what “done” looks like.
- Scope and a fixed price in writing, before any work starts.
- A working demo before the final payment. You see it work on your own data.
- Handover. You get the code, the access and a short guide to run it.
I use AI-assisted development tools, so changes ship fast.
Contact
Email is the best way to reach me: oleksandr@bratuta.net.
Tell me what you want to automate and which tools you use now.