Open to offers
Tech lead. I run the entire IT of two retail chains on my own, from 1C at the till to apps in the stores.
I design, set the tasks and own production. The code is written by a team of AI agents working to my process, so one person delivers the output of a whole department.
From the till to the app in the customer's pocket. I designed, built and run everything below myself.
Channels
Core
Data
Two chains in production with real customers, plus my own products.
Work · sports nutrition chain
Khabarovsk and Vladivostok. I built the chain's whole digital side: the app, loyalty bonuses, CRM, website, staff bot and the 1C integration.
Work · store chain and franchise
40+ stores in Khabarovsk. A loyalty program with 54,000 users: apps, prizes, tiers and discounts applied right at the till.
You just tell the CRM agent what you need: build an audience, launch a campaign, add product photos. It has an employee's permissions, and a human approves anything risky.
2 CRMs with MCP agents
In-store cameras count visitors and see whether a salesperson came over. The report lands in Telegram on its own.
43 ms per frame, down from 78
A shared support inbox for both chains: Telegram, WhatsApp, MAX and the website in one window, with AI-suggested replies.
410 tests in CI
My product
A team of AI agents on your Mac and server. A Rust daemon with native apps for Mac and iPhone, 20 releases in the first week.
My product
A messenger with Telegram, WhatsApp and its own chat in one app. Swift, TDLib and a Rust server, about 400 tests.
From web analytics, 1C and logs. No company revenue figures.
×4
SEO used to be run by a paid agency. Since October an AI agent does it: articles, fixes, reports. Reservations 21 → 87, visits +18%.
+300
Found in the data why the chain's receipt count dropped: free add-on deliveries to regular customers had stopped. They came back, and so did the receipts.
1.8×
78 → 43 ms per frame, 96 frames per second on a single GPU instead of 70.
120×
60 GB → 508 MB. Tracked down an endless retry of hopeless jobs: 54 million identical errors.
−58%
419 → 174 of 1,308. An agent found photos on manufacturers' sites; the rest went to the team as a shot list.
20
Every change is a PR, CI and a deploy. If the post-deploy check fails, production rolls back to the previous version on its own.
I'm the architect and I own production. A team of agents writes the code to my process, which is open source on GitHub ↗
I work through the task with the business. Money, bonuses and 1C get thought through before any code: transactions, ordering, rollback.
A senior model breaks the task into steps of 1–5 files, each checked by a single command.
Smaller models write code to the spec: tests first, then the implementation until it's green.
A separate agent reviews the diff, hooks block anything red, CI runs the tests.
Merge, and it's live. If the post-deploy check fails, it rolls back automatically.
BackendiOSMobileFrontend1CBitrixAnalytics / BIComputer visionAI engineerDevOpsServersMarketing / SEOSupportProduct and design
Remote or in Khabarovsk. I reply the same day, online 10:00–23:00 Khabarovsk time (UTC+10).
