Independent research
Analytics and research at massive scale, in parallel and from many sources.
I combine 23 years in IT and marketing with a network of AI agents, so you get one person working at the pace of a team. Brand, communication, strategy, AI, web, hardware. You pay for expertise and results, by the hour, on a B2B invoice.
Analytics and research at massive scale, in parallel and from many sources.
Strategy, solution architecture, code and linguistic nuance.
Independent oversight and cross-verification of outputs.
Engine: swappable frontier models (currently Claude Opus/Fable, OpenAI GPT Sol, Kimi K), chosen per task.
B2B invoice, billed in whole hours, with one hour as the smallest line item.
Ongoing team support and quick consultations.
In-depth work on documentation and systems audits.
Brand, message, technology. The choices you cannot take back.
Tick what you actually need. You'll see the hours and the cost before anyone sends you a quote.
Tick at least one item and I'll do the maths.
Indicative ranges, meant for a conversation. They're not a commercial offer or a fixed price list. A call usually narrows them, and sometimes it turns out half the items are unnecessary.
Send this scopeAn ongoing partnership for companies that want one accountable person on the other side. I cover both layers, the brand and the technology. I shape what you say to the market, and I stand behind the system that has to deliver on it. Nothing gets bounced back and forth between two vendors.
Deploying AI well is an architecture problem, and good architecture stops errors before they get out. My own approach cuts hallucinations by up to 80% through multi-stage, cross-model verification of the output. See the method in practice: the "One Question, 91 Minds" experiment (in Polish), an open workflow free to download.
Multi-stage verification and cross-model checking, up to −80% errors.
900M+ tokens daily in production, API cost optimization.
Safety-First architecture, GDPR compliance.
"A hallucination isn't a bug in the model. It's a failure in the architecture that let it through."
14-day VAT invoice. Minimum unit: 60 min. Copyright transfers to the client.
Response within 24h on business days. Urgent mode (6h) +100%. Remote work, Polish as the working language (EN documentation).
Mutual NDA on request. Enterprise scale: projects generate 900M+ tokens/day, with the API budget on the client's side.
I don't ask IF something will break, only WHAT happens then. Every layer of my ecosystem has a backup, an alert, and a recovery plan. Client projects get the same standard.
Every project lives in a private Git repository (GitHub). Full change history, so any state can be restored, and the history goes to the client along with the code.
Automated server snapshots (Hetzner) + daily copies to independent storage outside the main infrastructure + local copies. Three copies, two media types, one always offsite.
A standby server at a different provider, replicated live. If the main one fails, you lose seconds of data and you are back online in a few minutes.
External uptime monitoring plus internal health checks for services, certificates, and backups. I catch problems before a user reports them. Monitoring and alerts run automatically 24/7; human response follows the agreed hours and SLA.
Every anomaly — from an expiring certificate to a failed backup — lands instantly as a push notification (Pushover). Even at night.
When a project needs it, I put it behind a Cloudflare proxy: DNS, attack protection (WAF), and origin hiding. Traffic is filtered at the edge before it touches the app.
The ones you're already asking yourself, so here are the answers.
Documentation starts on day one, source files and access credentials stay with you, and the architecture has no vendor lock-in, so any competent team can take over the project. A retainer adds an SLA with a defined response time.
Yours. Copyright transfers to the client (standard practice, see Billing), documentation in Polish, technical docs also in English.
No overhead: you don't pay for coordination, handoffs, or subcontractor margins. One person owns the decision and the outcome, and the network of AI agents gives you a team's pace. An agency makes sense when you need a long-term team for years. If that's you, I'll tell you straight.
We define the criteria before we start: process time, cost to serve, error rate. We measure in production, not in a demo. And if there's no business case for AI at your company, you'll hear that before the first invoice.
Tell me the project status, the expected hourly budget and the timeframe. I'll come back with a concrete proposal for the next step.
Contact form