What AI runs on here and where it's headed:

• Triage of Error Logs from Test Runs
• AI-powered automation of unit updates
• AI Analysis of Race Results
• A custom tool for the sales team that replaces a CRM and maps the market. No spam bots.
• DGX Spark as a lab—internal applications and fine-tuning. We’re preparing a whole fleet of them, led by significantly more powerful hardware, and on top of that, training on a supercomputer.
• Knowledge Transfer with a Technical University
• There are plenty of other internal AI applications that we can discuss in person

What specifically will you be dealing with:

• Existing products / or new ones built from scratch: design, development, and operation of AI products and agent-based systems through to production
• Agent-based architecture: agent loops, multi-agent graphs, tool calling, RAG, MCP, multimodal models
• Custom Models: Fine-Tuning and Optimizing the Performance of Open-Source Models for Specific Use Cases
• Production Quality: Evaluation, Observability, Latency and Cost Management, Security, and Compliance
• Technical Decisions: We’ll involve you in the discussion about system design. Ego takes a back seat—quality comes first.

Our environment (we do not specify a language):

• LLM Ops: LangGraph, LangFuse, vLLM, PEFT, Unsloth, LiteLLM
• Infrastructure: Ubuntu, Docker, GitLab CI, Coolify, Fireclaw, M365
• Languages used by the team: Go, TypeScript, Python, Rust, Bash
• Coding agents: Codex, Claude Code, OpenCode

What you need to know:

• We don't count years of experience—we're interested in what you've actually built.
• At least one LLM system that was running in production and had users. Not a PoC, not a hackathon.
• Distributed Systems Architecture – Good Technical Decisions Regardless of the Language
• Adaptability: In a single afternoon, you can switch technologies, set up a benchmark, or create a comparative design in two languages
• AI-native workflow: You use coding agents to get up and running with a new stack in hours, not weeks
• You can make technical decisions off the top of your head: assess when it makes sense to use MCP and when to use Skills + API—and be able to explain why
• We would really appreciate: fine-tuning and self-hosting (vLLM, PEFT, Unsloth), custom agent graphs/loops, evaluation
• Czech/Slovak at a native-speaker level
• Advanced English – we deal with international clients and vendors

Why choose us:

• AI deployed in real-world engineering processes, not an internal chatbot. Proprietary MCP or RAG.
• Our own hardware and self-hosted models, not just calls to third-party APIs
• Your opinion and experience matter—you’re the third member of our new business unit—the architecture, tech stack, and processes are still taking shape, and you’ll be helping to shape and influence them as well.
• Collaboration with "AI celebrity" Michael Petrik 😉

We are looking for someone:

• Someone who takes a holistic approach to the system—its status, evaluation, fallbacks, and infrastructure—rather than just the prompt and the model.
• Anyone who doesn't mind that there are only two of us for now. That means a lot of decisions will be up to you.
• Who can handle the fact that part of the job involves maintaining and fine-tuning what’s already up and running—not just building something new?
• Who can think critically and make decisions

Send us

• a link to the code (GitHub, a sample, anything you've actually built)
• In a nutshell: What AI system did you put into production, and what went wrong with it?
• Resume/LinkedIn

‍Office: Mladá Boleslav – at least 2 days a week in person; the rest can be remote work

Type of employment: Full-time/part-time; Full-time employment contract/Self-employed

Start date: to be agreed upon