„We're planning a new system. We don't know what to build, or what it costs.“
You need to settle scope, architecture and budget before you approach vendors — and be able to judge what they quote.
02 / Where it comes from
My track record in three layers.
This doesn't come from slide decks and theoretical frameworks. I worked each layer in production, where mistakes have real financial and operational consequences.
Layer one — Enterprise
Where I learned what "it must not go down" means.
A recruitment system for PPF Group. A warehouse system for Toyota Tsusho. Leading the frontend team for internal applications at Komerční banka. Environments where hundreds of people depend on the system, where you go through audits, and where a bad architectural decision is paid for over years. That's where I learned that technical elegance is the second question, not the first.
Layer two — My own team
Where I took responsibility for keeping it running.
I founded Techcrowd, today around fifteen people. For Brink's we replaced the legacy cash processing system — it has been running for four years. For STING Reality, an intranet and a mobile app for a network of estate agents. For ABC pack, warehouse control with a real-time view of pallet movement. The difference from layer one: here my name is on the operation, not just the design.
Layer three — AI in production
Where I separate what works from what demos well.
For the past few years I've been building AI layers on top of company data — RAG over documents, vector search, custom agents, MCP integrations. Not a demo for a presentation, but systems with an audit trail, controllable cost, and someone accountable for them. Most companies that call me already have a pilot that never made it into production.