Phone
Address
Wehlistraße 291/1/47
1020 Vienna
DIE WERKSTATT · VIENNA
Your company can build its own software now
It still needs a foundation. I'm Michael Wallner. I lay the groundwork, set up an AI development workflow your team can actually operate, and hand the whole thing over. Then you keep building without me.
Proof, not promises
Four products. One person.
b10cks, a headless CMS. RAVN Mail, a cross-platform email client. Kessel, a remote file manager in Rust. gh0st, a terminal SEO crawler. All open source, all shipped by one developer using exactly the workflow I set up for client teams. The method isn't a theory I'm selling. It's how this site got built.
Code got cheap. Systems did not. That's where I come in.
The foundation
Data model, permissions, environments, deployment, and the boundaries everything else hangs off. The decisions that are cheap to make once and expensive to unmake later.
The workflow
Agents, steered properly. Instruction files, MCP integrations against your systems, spec-driven loops, evals and guardrails, review gates. The output stays reviewable by people who aren't engineers.
The handover
Then I leave. Your people are in the repository from week one and shipping by the end. Documentation, runbooks, and training included. No retainer you can't escape.
Four Phases
Two to three months, from empty repository to a team that ships without me.
1. Foundation
Architecture, data model, authentication, CI/CD, deployment. At the end there's a real application in production. It just doesn't do much yet.
2. The Golden Path
One feature built properly, end to end, as the pattern every later feature copies. It's also what your agents imitate, so it has to be right.
3. AI Development Workflow
Context engineering, MCP integrations, sub-agent roles, eval harnesses, and guardrails. Plus loop and graph engineering for the multi-step work where a single prompt drifts.
4. Handover & Enablement
Pairing, runbooks, and practical training in steering agents. For developers if you have them, for capable non-developers if you don't.
Craft, in the Age of AI
The workshop metaphor is deliberate. A master craftsman's real legacy was never the objects. It was the apprentices who could work without him. That's the model here. Your team are the apprentices, and the agents are very fast, very literal-minded hands.
Judgement is the scarce part
A model will write whatever you ask for, confidently, including the wrong thing. What it can't do is decide what deserves to exist, where the boundaries belong, or which shortcut will quietly cost you a year.
That judgement is what fifteen years buys, and I spend the engagement transferring it: in the structure of the code, the constraints in the workflow, and the conversations we have while we build.
Built to be handed over
Every project ships with documentation, tests, and a deployment pipeline your own team can run. I write code assuming someone else will maintain it. In this model, someone else definitely will.
The repository, the pipeline, and the decisions are yours from the first week. An engagement that ends cleanly is the goal, not an outcome to be avoided.
Open by default
The tooling underneath is open source: b10cks, SabaccUI, Kessel, gh0st, RAVN, and numerous open source packages on Packagist or NPM.
You can read all of it before you hire me, and you're never dependent on a proprietary layer you can't inspect. If you later hire developers or move to another partner, that transition should be boring.
Who This Is For
Companies without a development team
Manufacturing, logistics, healthcare, professional services. You know your process better than any vendor, and the standard product still doesn't fit it.
Founders without a CTO
You understand the market. You need a foundation that survives early traction and the ability to keep iterating while you find the right team.
Teams adopting agents
You have developers but no workflow. Quality swings, reviews are guesswork, and half the experiments never merged. The workflow phase can be booked on its own.
Anyone whose AI pilot stalled
The demo worked. Then came authentication, permissions, migrations, deployment, and no way to tell whether the output was correct. That gap is the job.
Based in Vienna, Working Everywhere
In the room where it matters
Vienna-based, with regular presence in Salzburg. The foundation and handover phases go faster in person. For Austrian and German clients that's usually a few days on site, not a permanent desk.
European by default
GDPR-compliant architecture, EU data residency, and a clear answer to which model provider sees what. Decided while building, not audited afterwards.
Ready to start?
Bring the problem, not the specification
Thirty minutes, no pitch. Most first calls start with a process held together by spreadsheets and goodwill. That's a perfectly good place to begin. You'll leave with an honest read on whether this is worth building, and what it would take.