No-code developer on Make and n8n. I build Telegram bots, RAG systems, and integrations between services — from a constructor prototype to self-hosted on the client’s server. When a scenario is not enough, I add the missing piece in Python: site and PDF parsing, Excel/Word export, desktop utilities.
I came to no-code automation from game development in C++/Unreal Engine. I build Telegram bots, RAG systems, and integrations on Make and n8n. Self-hosted setups (Docker, Linux) run on the client’s server — or, as a separate service, I can provision and configure a server from scratch. Make and AI model fees are agreed separately; they are not included in the development price.
When a constructor has no matching step, I ship a working Python tool — not a throwaway script: a GUI, logs for doubtful rows, writes into the client’s files. I write code with an AI assistant: I set the task, check the logic, test on real files, and deliver what I can support. Same studio habits from C++/Unreal: decompose, stage, ship.
Not client releases — systems you can open and try. Each one solves a typical business pain; the same pattern can run on your data. The public bots are in Russian; production builds can run in English.
Free text → spreadsheet rows: you write as you speak (“coffee 350, salary 80,000, taxi yesterday 1,200”) — the bot distinguishes expenses from income, fills date, category, and amount, and appends a Google Sheet. The sheet totals expenses and income and has a tab per month. Anti-spam and a daily cap fire before the paid AI call. Two storage modes: a fast shared template or a sheet on the user’s Drive.
The same pattern works for inbound requests and chat parsing, not only personal finance.
Open in Telegram → @MyFinanceNotes_botMulti-step dialogue → a ready document: you answer step by step, the session lives in PostgreSQL, the output is a claim file in Telegram. I do not store personal full names in the database: the person fills them into the template themselves. Stack in Docker on Linux (the demo runs on my VPS; client builds run on your infrastructure, or on a turnkey server as a separate service).
The same skeleton works for intake forms and template documents, not only complaints.
Open in Telegram → @shaloby_helper_botAnswers from a knowledge base and a catalog — without making things up. Meaning questions (delivery, returns, warranty) go to vector search; exact ones (“how much”, stock) go to SQL on products and prices. If neither source has an answer, the bot does not hallucinate: it logs the question for a live manager and says so.
Open in Telegram → @TechStorePro_demo_botI keep the constructor where it covers the job. I build a separate Python tool when that is faster and more reliable: messy PDFs, a site with no proper API, a desktop app with a window and a log — or when I need custom AI logic that would take longer in Make/n8n. Two working GUI apps below.
A desktop utility for messy legal PDFs: a folder of arbitration acts plus a register of practitioners → rows in a working Excel file (16 fields). Writes into an existing file by case number; .xls is saved alongside as .xlsx. Empty fields and dismissal acts are highlighted; it writes a “check this” log — no silent failures.
The hard part is not “just regex”: floating wording, broken PDF text, names after organization IDs, maiden names in parentheses, the act date rather than the filing date.
A desktop app for insolvency-practitioner SROs: collects debtors and procedures from a public register, keeps an internal project (JSON, no duplicates), exports Word. SRO templates are profiles; the app does not overwrite other people’s acts. Collection via Playwright, manual row edits, an explicit log of data gaps.
License and trial period via a small API on a VPS (Docker, HTTPS). Debtor names and cases never leave the machine — only key accounting goes to the server.
You describe the task as it is — goal, constraints, what already exists: data, services, access.
I write a spec: scenarios, integrations, data. I add what usually gets missed at the start, and split the work into stages with dates.
We agree the spec, stages, and budget. I start building only after that — so the scope does not creep mid-project.
I assemble a working prototype on Make, n8n, or Python if that closes the core logic faster.
I cover edge cases, limits, and replay protection. Make runs in the cloud (plan billed separately). n8n runs on the client’s server, or on a turnkey server as a separate service.
I hand over access and a short write-up, and stay on support if needed.
Tell me what you need automated — I will reply with a timeline estimate and an approach.
Telegram: @AndrewUDV
Email: andrewdovich88@gmail.com