I bring AI into businesses that already run — e-commerce, clinics, service companies of 10–200 people: agents that answer clients, voice automation, document processing, marketing-ops. Everything I build for clients, I run in my own production first.
From the first customer touchpoint to your own app — production systems with measurable payback.
This is my own battle-tested stack. The same approaches go into your business: here are the exact building blocks I'll assemble your solution from.
Campaigns are assembled and launched over API, creatives are generated automatically, conversions flow into the tracker, and every morning a summary of yesterday lands in Telegram. Running daily on live campaigns.
Five production servers in Europe. Services under systemd with auto-restart, centralised logs, nginx with certificate auto-renewal, an encrypted tunnel between nodes, watchdog scripts and alerts to Discord and Telegram.
Several of my own systems making decisions in real time: every few seconds they look at incoming data, act, write everything to a database and alert to Discord. Fintech, domain under NDA.
A database recording every meaningful action across all my projects, plus a knowledge graph for long-term agent memory. Scheduled jobs keep it consistent and significant decisions are mirrored to a dedicated Discord channel. This is why an agent still has the context months later.
A step-by-step read on a competitor from public data: who is behind the site, what it is built on, how the traffic chain is put together, what changed over time. Legal public sources only — registration records, certificates, page archives. Scripted, takes 30–60 minutes.
Two independent browser profiles run in parallel and survive environment updates: a background service restores the configuration on its own, with the rollback recipe kept in the repo. Every update used to cost half a day of manual repair.
Every system is my own code. Deployed on my own servers, under my monitoring. This is my personal production stack — I don't disclose the specifics of domains or clients, but I carry the same architectural approaches into client projects.
Fixed quote before the work starts, a progress recording every week, access to the working version from day one. You see how it comes together, week by week.
A call over Zoom, Telegram or Google Meet — whatever suits you. We walk through how things run today and I map out 3–5 points where automation pays for itself in 1–3 months. If I don't see the payback, I say so and we stop there.
If we move on, I write a detailed spec: what exactly we build, on what, how it plugs into your CRM and accounting. The quote is fixed — the number in the contract does not drift mid-project.
Once a week I send a short screen recording: what got done, what's next, where it got stuck. You get access to the working version straight away and try it out as the work goes on.
I keep it running: monitoring, updates for new scenarios, moving to better models as they ship. I reply within 2 hours during business hours.
Examples from work that isn't under NDA. There is more under NDA — I'll walk you through it on a call.
The accountant processed Treasury XML statements by hand to load them into 1C — 2–3 hours per statement, many times a month.
Built a parser for both versions of the Treasury protocol (V3 + V4) with conversion into the 1CClientBankExchange format (Windows-1251).
The team spent 4–6 hours a week assembling reports: FB Ads + Yandex.Direct + Metrica + creatives across 8 clients.
An n8n workflow: it pulls the numbers from every ad platform, AI writes the commentary on anomalies, the report is assembled in PDF or Notion and goes out to the client Monday 9:00.
Launching 50–100 campaigns a day by hand, plus hunting for competitor creatives — 20+ hours per buyer every week.
A Python tool: pre-flight checks + campaign launch via the FB API + automatic parsing of the FB Ad Library for the top competitors in the niche.
"I build systems that still work a year later."
My name is Vyacheslav. In IT since 2017. I started as an engineer at the Baltic Shipyard — shipbuilding, Primavera, multi-level schedules across 10,000+ material items. Then Infor LN, BAAN 4gl, ERP customization for foreign clients in Turkey (IPL Consulting) and Bulgaria (Asertiva Solutions, EU market) — clients included Ferrari, Aston Martin, Kozloduy NPP, Rosatom, OSK + plants across RU / EU / the Middle East.
In parallel I kept studying: App Brewery — iOS & Swift Bootcamp, then Web Development. Shipped 20+ iOS apps (Swift, SwiftUI, CoreML, ARKit) and took part in WWDC21.
In 2024 I moved into media buying: a large media-buying team (Middle), then Senior at a second team. Both names are under NDA. $1.34M of ad budget in performance marketing, $379k net profit over the period. That is also when I started building my own AI tooling: auto-launch of campaigns via the Graph API, creative parsers, lead classifiers on GPT.
Today — my own AI agents running in production (trading, growth, automation) and bringing those same approaches into SMB. Working systems, end to end: code, launch, ongoing support.
All figures are starting prices: a simpler task costs less, a bigger one more. The final number is fixed in the quote after the written spec (free).
Where to start. We go through your processes on a call and you get a map of automations with a payback estimate.
One process end-to-end. We prove it on a specific task before committing to a bigger system.
The full system: several channels at once, integrations with your CRM, team training.
After launch. I keep it running and extend it as new tasks come up.
I work under a civil contract as samozanyatyy (self-employed). For larger contracts — engagement via a legal entity.
What an implementation costs, what a booking bot is actually made of, and what the law lets you send to an LLM. Numbers and caveats from my own projects.
Six quotes from my own projects: a ₽600K process audit returned in 7 months, ₽480K for 1C access via Telegram, an AI security check from ₽45K. What each includes.
ReadWhat an AI receptionist and booking chatbot is made of, the four places it breaks in the field, and nine questions to ask a vendor before you pay.
ReadCan you send customer data to ChatGPT, what 152-FZ requires, why most anonymisation doesn't hold up, and how to audit your own system in 20 minutes.
ReadI'll tell you what to deploy in your business first, what the payback looks like, and whether you need AI for the task at all (sometimes you don't).
Or just send your question — I reply within 2 hours