Skip to content

Kraków · working globally · 2 slots open for Q2

AI systems that actually ship.

We build, rescue, and amplify AI for companies that have moved past "should we use it" and need someone to make it work.

Four productised services. One consulting practice. Based in Kraków.

Now serving: Recipes — 30+ skills for your agent →

Most AI projects fail between the prototype and production. The model works; the system around it does not. Integrations break. Agents hallucinate on real data. Nobody owns the monitoring.

WiseChef works with companies that have already decided to use AI and need it to actually function — not a proof of concept, a running system. We build it, we fix it, or we tell you what is wrong.

Adam Krawczyk founded WiseChef after years of building AI systems at the edge of what tooling supported, including computer vision work through WiseVision. That background shapes how we work: infrastructure first, abstraction second, honesty throughout.

Products

Four ways to work with us

How we work

Four steps. No mystery.

The same process runs behind every product and every consulting engagement. It is the reason things ship.

01

Diagnose

We look at what you actually have — configs, agents, data flows, monitoring — before proposing anything. No generic audits.

02

Scope

We write down what we are going to build or fix, what it costs, and how long it takes. You approve before work starts.

03

Ship

We build and deploy into your environment, on your infrastructure where that matters. You can see the code at every step.

04

Hand off

You keep the system, the documentation, and the knowledge to run it. Optional monitoring retainer if you want ongoing coverage.

Consulting

Need something custom?

A full agent rollout, a failing integration diagnosed end-to-end, or ongoing monitoring across a multi-system AI stack — that is what the consulting practice is for. Engagements from €2,500.

Background

Built on the same stack we sell.

WiseChef architecture: 4 layers — Client access, Agent layer, Data layer, Infrastructure

Infrastructure

Hetzner Cloud across three EU regions. Cloudflare tunnels and DNS. Postgres, Redis, and Elasticsearch for agent state. Docker for isolated per-tenant services. Everything the Framework product provisions is the same stack we run internally.

Agent layer

OpenClaw for conversational runtime. Cognee for structured knowledge and retrieval. Paperclip for task orchestration across agents. Multiple LLM providers behind a fallback router so a single outage never takes us down.

Founder

Adam Krawczyk — software engineer, founder of WiseVision (computer vision and robotics, Kraków). Years of production ML work before language models became a category. WiseChef applies those engineering standards to AI agents and the infrastructure around them.

Engagement model

Fixed-scope engagements. Plain-language invoices. Code you can see and keep. No white-label reselling of someone else's platform. No locked-in abstractions between you and your data.