I build AI systems that take routine off your hands and lift your growth ceiling — Claude Code workspaces, agents, integrations, private AI environments. For teams and departments that need AI integration and development.
Discuss a projectMy name is Vlad. I work and create under the name YUSU.
Marketing for real: funnels, CRM, landings, paid traffic, SMM, copy, course launches. I did this long before "AI marketing" was a buzzword.
In a parallel life — musician and artist. DJ and live guitar, electronic music, generative media art enthusiast.
I like informal, clear, transparent communication.
Currently based in Thailand. Working in a stack of Claude Code, Framer, Figma. I do everything with my own head and hands. And the hands of AI agents.
YUSU.lab is the lab where I build client systems the same way I build my own — with the same involvement and accountability. I work in mine every day.
I do one of two things: take manual routine off people's hands — the process starts costing less, or build a system that removes your growth ceiling — the team gets more done without growing headcount. We count in money: what the system saves or brings in.
Light (5-7 days) — clean template, branded, when speed matters. · Custom-brand (10-14 days) — designed from the ground up, a face of your own. · Redesign (7-10 days) — you have a site, but it doesn't say who you are; rebuilt without losing content or SEO.
Example: I built a four-page training site for a cacao store, and brought six websites of two clients in line with data law: consent in forms, policies, banners.
With a conversation. You tell me about the task and your processes, and I tell you honestly where AI pays off — and where it doesn't. Data first, conclusions second.
Case studies from real work — numbers and stack from shipped projects and my own system, not promises.
So far I've worked with small teams and brands. I haven't had a mid-size or large company as a client yet. I'd build your team's system the same way I built my own: I work in mine every day, and I'll show it on a call.
All my work sits in one repository — from client projects to music. The workspace has memory: notes on how I work and which mistakes not to repeat. The rules live there too, like nothing gets published without my OK. So do my own skills: ready-made instructions for tasks that keep coming back.
Some of the work runs without me. When I sit down, the day's brief is already there. On Mondays I get a summary across all projects, and every day a digest of Telegram channels. In early October ten agents cleaned up the whole repository, and I only made the decisions.
A small firm wanted its own agent on its own server — always on, and reachable from a phone.
I put Claude Code on a dedicated server and locked it down. From a laptop people connect through their code editor, from a phone they type or dictate in Telegram. Memory persists between sessions. Each person has a separate user with their own Claude subscription and history.
A handmade designer clothing brand from Southeast Asia. Strong in person — but the online store barely sold: the catalog had grown by hand over years, product cards kept breaking, and shoppers from other countries saw the wrong prices in the wrong currency and couldn't check out. That was point A.
I rebuilt the store under one standard, on my own with AI. I put the catalog in order and set up prices and shipping for Thailand and 50 European countries.
Two clients, three sites each: a manufacturing company in Russia and a bureau in Germany with an online shop. Forms collected data without consent, analytics and ad trackers fired before it, and the privacy policy was outdated or missing.
I added consent to the forms, prepared policies and legal pages in Russian, German and English, and set up consent banners. One old site stopped collecting data altogether.
An online cacao store on Shopify. It paid every month for review and subscription apps, and training applications lived in chat threads, tracked nowhere.
I moved the reviews to a free service and the subscriptions to Shopify's built-in ones. For the training I built a separate site: an application lands in the store by itself, and after paying by link the person gets a confirmation.
The first working system — 1-2 weeks. A site — from a week. Exact scope and timelines are locked after the first conversation and the brief. Basically, the first result lands within two weeks.
USD / EUR / USDT. A job with clear edges — 100% upfront. A large one — 50% upfront after scope is confirmed, 50% on delivery. No kickoff without upfront.
2 rounds of revisions included on any service. If something doesn't fit after scope sign-off — we rework within those rounds. The upfront payment is non-refundable after kickoff (it covers my reserved slot).
From an agency — one person on the project, no juniors or middle managers. From a subscription AI tool — strategy and a setup tuned to your tasks, data and rules, not generic ChatGPT. And real marketing experience behind it.
The workspace lives with you — on your machines, in your repository or on your server. There's no server of mine in the middle. The company takes out the Claude subscriptions in its own name. Whatever the agent reads while doing a task goes to Anthropic for processing — we go through your plan's terms together before we start.
That's fine. I'll help shape it during the kickoff conversation — that's part of the work, not a separate service.
Hit "Discuss a project" (top of page) or email yusulabs@gmail.com. Brief is 30 minutes, reply within 24h.
Write a few lines in the form — what team, what task. Or email me directly, reply within 24h.
A few lines free-form — what's the project, the budget, the timing. I'll reply within 24h to your email or WhatsApp.