# Ozan Atmar

> Full text of ozan.at, inlined for language models. Ozan is an automation and full-stack contractor based in Varna, Bulgaria, available for on-site work across the EU and remote project work worldwide.

Source pages: https://ozan.at/mar and https://ozan.at/mar-training
Contact: ozan@ozan.at · book a call https://cal.com/ozanatmar · LinkedIn https://www.linkedin.com/in/ozanatmar/

## Portfolio (https://ozan.at/mar)

Ozan Atmar. Automation and full-stack contractor based in Varna.
I build the workflows, integrations, and small apps that founders need shipped this month.

Stack: Make.com, n8n, Zapier, Next.js, React Native, Node, Python, Supabase, OpenAI, Stripe, Vercel.

### Projects

#### Atmar Horeca (https://atmarhoreca.com)
B2B e-commerce for hotel, restaurant, and café equipment. EU-wide delivery, VIES VAT validation, 16-page admin.
Stack: Next.js, Supabase, Stripe, OpenAI, Brevo, Make.com. Role: solo. Status: live.

Atmar Horeca sells professional equipment from European brands to hotels, restaurants, and cafés across the EU. B2B e-commerce carries rules consumer platforms skip: VAT numbers validated against VIES, zero-rated cross-border invoices, large orders that need a human decision before payment, and suppliers with no stock feed. Off-the-shelf shop platforms handle some of that. The interesting parts they don't.
Guests browse and build a cart, ordering requires an account. Standard orders pay by card through Stripe. Larger orders go to admin approval: approving one computes shipping from route and handling-day data, revalidates the VAT number, creates a Stripe Payment Link, and emails a proforma invoice generated with pdfkit. The order API ignores every number the client sends and re-prices from the catalog, and the Stripe webhook checks amount and currency against the stored total before anything is marked paid.
EU VAT took the most care. VIES, the EU's VAT validation service, reports member-state outages in a way that looks like an invalid number, so the pipeline treats every ambiguous answer the same way: zero-rate provisionally, force the order into admin approval, flag it for manual review. Nothing is invoiced or charged until a person verifies. Each order snapshots its full tax state at creation, so changing a customer's VAT number later never rewrites an issued invoice.
The admin panel is 16 pages: an order state machine, product forms with option groups and price modifiers, a support inbox, and a bulk pipeline that round-trips the catalog through Google Sheets, validating every row before inserting any. Supplier imports scrape the product page, let GPT-4o rewrite the content, and pull per-variant images matched by SKU. The launch supplier has no stock API, so a scheduled job diffs their sitemap against stored product URLs to infer stock. When the scrape fails, products flip to 'On Request' rather than showing a stale 'In Stock'. Built solo in two months on Next.js, Supabase, Stripe, Brevo, and Make.com. Live since spring 2026.

#### Namedrop (https://namedrop.cc)
AI business name generator for founders. $9 one-time, used in 40+ countries.
Stack: Next.js, Supabase, OpenAI, Stripe, Resend. Role: solo. Status: shipped, ongoing maintenance.

Founders waste real time on naming. They check domains across registrars, search trademarks, test how a name sounds, and end up with a shortlist that's mostly variations of the same word. Most AI name generators make this worse by returning long lists of obvious combinations without filtering anything.
The user describes their business in a few sentences. The pipeline sends that to an OpenAI prompt that returns ten names with short rationales. Each name is checked across multiple TLDs via the Porkbun API, with .com availability as the strongest ranking signal. For trademarks, OpenAI classifies the business into a Nice classification code, then the USPTO API checks each name against that category and returns one of three statuses: clean, conflict in another category, or direct conflict. Results are emailed via Resend.
Built in two weeks. Next.js on Vercel, Supabase for storage, Stripe Checkout for the $9 one-time payment, OpenAI on the server. No accounts, no subscriptions, no dashboard. The user pays, gets ten names, and leaves.
Live since 2024. Used in 40+ countries.

#### Truvex (https://truvex.app)
Native iOS/Android app for restaurant shift coverage. Twilio SMS dispatch with Stripe billing.
Stack: React Native (Expo), Supabase, Twilio, Stripe, OpenAI, Next.js. Role: solo. Status: pre-launch.

Restaurant managers handle shift callouts through personal texts and group chats. When someone calls out at 6am, the manager works through a mental list, hoping someone picks up. The bad version of this is one shift uncovered. The worse version is the manager covering it themselves.
Truvex automates the chain. The manager submits a callout in the app, Twilio sends an SMS to every team member whose roles match the open shift. Employees claim the shift in one tap. The manager picks one of the applicants. If the manager doesn't pick within 30 minutes, the first applicant gets the shift automatically.
Built solo. React Native (Expo) for iOS and Android from one codebase. Supabase for auth, team rosters, and shift data with row-level security. Twilio for inbound and outbound SMS via webhooks. Stripe for per-location subscriptions, with Pro and Business tiers including SMS and push, Free tier without.
The marketing site has its own automation: a single OpenAI pipeline that generates a blog post, the cover image, and the social posts that link back to it. Three calls, one workflow, zero manual writing.
Pre-launch.

#### SuppliMatch (https://supplimatch.com)
B2B matching platform connecting supplement brands with contract manufacturers. Email-first, no signup walls.
Stack: HTML, Make.com, Notion, OpenAI. Role: solo. Status: live prototype.

Contract manufacturing for supplements runs on trade shows and warm intros. Small brands send cold emails to generic info@ addresses and get ignored. Manufacturers have open capacity but no qualified inbound. The matching problem is solvable, but no one wants to build a dashboard product when the actual workflow is an email.
SuppliMatch is a prototype I built to test the hypothesis. Brands submit a brief through a form. Make.com workflows route the brief to manufacturers whose registered capabilities match. Manufacturers receive the brief by email only. They reply with a quote, the brand reviews quotes, and contact details are revealed only after the brand accepts. No accounts, no logins, no dashboards on either side.
Stack is intentionally light. Notion is the database. Make.com runs every workflow, including signup verification, request handling, matching, email dispatch, and the contact-reveal flow. The frontend is plain HTML and JS. AI is used in one narrow place: filtering submissions that look like attempts to source controlled or illegal substances.
Live. The point of the build was to test how cheap and fast a real workflow can be before deciding whether to build a coded version. Most of the time, the prototype is the product.

### Contact
Send a brief. ozan@ozan.at
Book a call: https://cal.com/ozanatmar
Currently taking on 1 to 2 new projects per month. Reply usually within 24 hours, weekdays.
Good fit: defined problem, named budget, founder or operator decision-maker. Not a fit: unfunded ideas, long retainers, recruiter pipelines.

## AI automation training & mentorship (https://ozan.at/mar-training)

I teach your team to automate the manual work that eats their days, then stay on as their mentor.
If your staff spend hours every week on invoices, data entry, reports, and follow-ups, most of that can run itself. I come to your office, train your people to build the automations on your real tasks, and stay on as their mentor. Not a webinar you forget by Friday. The goal: handle more work without hiring more people, and free your team from busywork.
Based in Varna, Bulgaria · Travel across the EU · In English, on-site

### What I actually teach
You don't need to know what Make.com or n8n are. By the end, your people will. Two core tools, taught properly, with AI wired inside them. Not a tour of forty apps you'll never open. The goal is that your team can build and fix automations on their own once I'm gone.

Tools I teach:
- Make.com: the main teaching tool, visual and fast to learn
- n8n: the power tool, self-hosted, your data stays in-house
- AI in workflows: ChatGPT and Claude wired into the steps
- Zapier: covered if your team already runs on it
- APIs & webhooks: the glue, taught in plain language

This is for you if:
- You run an accounting firm, shop or online store, warehouse, clinic, workshop, agency, or similar, and your people lose hours to repetitive manual work
- You want to handle more orders and clients without adding headcount
- Nobody on your team is technical, and you don't want to depend on an agency
- You'd rather learn face to face, in your own office, on your own tasks

Probably not a fit:
- You're a tech, software, or marketing company that already automates in-house
- You want a pre-recorded online course to watch alone
- You want me to build everything and hand over a black box
- You're looking for the cheapest possible option

### Four ways to work together
Start with the foundations, go deeper with a hands-on build, or keep me on as your team's automation mentor. Most teams move from one into the next.

#### Foundations: AI for daily work
A half day getting your team comfortable using AI (Claude, ChatGPT) for the everyday work, before any automation. Drafting emails and replies, summarizing long documents and contracts, turning messy notes into clean text, answering routine questions from your own files and procedures.
- Hands-on with your team's real tasks
- Plain language, nothing technical to install
- Best first step if your team is new to AI
- The on-ramp to automating the work later
Price: From a half-day rate, per team

#### Intro Workshop
A half or full day with your team. Everyone leaves understanding what is possible, and with their first working automation.
- Live, in your office
- Real examples from your work
- First automation built together
- Best for awareness and quick wins
Price: From a half-day rate, per team

#### Hands-on Build (Most impact)
Two to three days where your people build real automations for their own tasks, with me beside them the whole time.
- Make and n8n, hands on keyboards
- Solves actual processes you run
- Your team leaves able to repeat it
- Working automations, not slides
Price: Quoted by days plus team size

#### Mentor Retainer
The part that matters most. When your team hits a wall building something real, I unblock them. Month to month.
- Ongoing help after training
- Calls, screen-shares, fixes
- On-site visits when needed
- Your safety net while they learn
Price: Monthly, fixed retainer

### Training that does not evaporate
Most corporate training is forgotten within a week, because nobody is there when the real problem shows up. I fix that by staying.

- Short call: Twenty minutes to understand your team and where the time goes.
- On-site training: I come to you. We build automations on your real, current tasks.
- Your team builds: They start automating their own work, with confidence.
- I stay on: Stuck? I show the way. The mentorship is the difference.

The mentor part is the whole point. Anyone can run a workshop. The value is that when one of your people is halfway through building something and it breaks at 4pm on a Tuesday, they message me and we fix it together. That is what makes the skill actually stick, and it is hard for a one-off trainer to copy.

### The kind of thing we'll build
Concrete, everyday office work, the stuff that quietly eats hours. Your real examples will be different. These show the shape of it.

- Inbox → action: Incoming emails read, sorted, summarized, and turned into tasks or replies automatically. (make, openai, gmail)
- Invoice & data entry: Documents read, fields extracted, and pushed into your sheet or accounting tool, with no manual typing. (n8n, ocr, sheets)
- Reports on autopilot: Weekly numbers pulled from your tools, written up, and delivered to the team every Monday. (make, openai, slack)
- Lead follow-up: New leads logged, qualified, and followed up on time, so nothing slips through. (make, crm, email)
- Internal AI assistant: A private assistant that answers from your own documents and procedures. (n8n, claude, vector)
- App-to-app sync: Two tools that do not talk to each other, finally kept in sync without copy-paste. (make, webhooks, api)

### How pricing works
The offer is fixed. The price is a quote, because it honestly depends on you. Every engagement is built from three simple parts:

- Day rate: per day of training on-site
- Team size: how many people in the room
- Travel: outside Varna, added at cost

Mentor retainers are a separate, fixed monthly fee. Tell me your team size and where you are, and I'll send a clear number, with no surprises.

### Varna first. The EU when it makes sense.
I'm based in Varna, and that's where I start. But there are planes to everywhere, and I travel across the EU for clients who want me on-site. If your team is in Sofia, Bucharest, Berlin, or anywhere with a runway and English-speaking staff, travel goes on the quote and we make it work. Between visits, the mentorship runs remotely, so distance never blocks the help.

### Contact
Tell me where the time goes.
One short call. Tell me what your team does manually all day, and I'll tell you honestly whether automation training is worth it for you, and roughly what it would cost.
Book a call: https://cal.com/ozanatmar
Email: ozan@ozan.at