about
Operations first, then data,
now I build
My name is Michal Antal, but everyone calls me Myky. I did not get into software through a computer science degree. I got there through six years of running operations, where every wasted hour had a name and a price.
That is why I ask how a business works before I build anything. Most badly built systems are not badly programmed, they are built on a poor understanding of what the business actually does.
AI makes writing code faster. It does not decide what gets built, and it does not replace that first conversation.
How I got here
2018 → 2022
Three hundred events and budgets with no undo
I ran over three hundred events, from small gatherings to international championships, with budgets up to forty thousand euros and teams of up to thirty people across four countries. At an event there is no "we will fix it tomorrow". Either it works at ten on Saturday morning or three hundred people are standing there looking at you. It taught me to plan ahead and to assume something will fail.
2022 → 2023
Bolt, where a change in logic raised people’s pay
At Bolt I handled courier fleet logistics. It looked operational but it was a data problem. Couriers were being sent across half the city for orders that were not worth taking. I reworked the dispatch logic to account for where a courier actually was and where it was worth going. Average courier revenue rose 72% in two months. Those are not numbers in a report, that is money specific people took home.
2023 → 2026
Enterprise data and receipts for six hundredths of a crown
Alongside management studies at the Prague University of Economics I consulted on IBM Planning Analytics at Adastra. When the system went down I rebuilt eight critical reporting forms in a single week, the ones the monthly close depended on. My thesis was on autonomous data extraction from accounting documents. Commercial tools charge around four crowns per document. Mine got below six hundredths of a crown, roughly seventy times cheaper. The same principle now powers the feature in Paving Planner that turns a photo of a sketch into a drawing.
2024 → now
My own products
Six finished projects. Two are used daily by real businesses, one is a paid SaaS in seven languages. I also guest lecture at the Prague University of Economics on AI agents for data extraction.
The tools I work with
I do not rebuild from scratch what somebody has already built better. These are the open tools I have wired into my work, with links to their source. They are public, so you can look.
- claude-seo ★ 14 000
A full SEO kit inside the editor. Technical audits, structured data, local search and optimisation for being cited by language models, not just Google.
- humanizer ★ 35 000
Strips the tell-tale signs of machine writing out of text. That is why nothing on this site reads like it came out of a generator.
- watermarks-remover ★ 2 100
Generators leave a trace in what they produce: invisible characters in text and provenance data written into images and documents. This tool finds and removes it. I run it over my own writing as a final check, so nothing stays in it that I did not put there.
- ponytail ★ 101 000
Makes the tool think like a senior developer who hates writing code that did not need to exist. Less code means fewer things that can break.
- n8n-skills ★ 6 000
Builds n8n automations that work first time. This is where the plumbing between systems comes from, the kind that stops people retyping data from one place into another.
- headroom ★ 66 000
Shrinks the volume of data going in for processing. In practice that means cheaper and faster work with no loss of quality.
Writes and publishes articles optimised for search and for language-model answers. Useful for clients whose site needs to keep producing content.
Not a repository but a service. It checks finished content and reports what it is missing to rank higher.
External audit of security headers: CSP, HSTS, Permissions-Policy, referrer, XFO. Every site runs through this before it ships; the target is grade A or A+, nothing lower.
On top of that I subscribe to Claude Max 20, the highest tier. Every month I spend more on tooling than some agencies charge for a whole job. That is why the results look the way they do. I do not cut corners on what I actually build your work with.
Education and certifications
The list is not complete and I keep adding to it. The engineering and electrical qualifications are not decoration: a good share of my clients are manufacturing and trade businesses, and it makes a difference when you can talk about their work without them having to explain it first.
Engineering trade
- Engineer of mechanical engineering
- Electrical qualification §21
- Flame welding
Data and AI
- Claude Ecosystem Mastery by Anthropic (MCP, Agent Skills, AI Fluency, Bedrock and Vertex AI)
- IBM Planning Analytics (TM1)
- Data Science Professional (KNIME)
- MS Excel Expert (NASBA preparation)
Communication and leadership
- NLP Practitioner & Master
- Professional speaking (Toastmasters)
Languages
- Slovak · native
- English · C1
- Spanish · B1