Job Description

We’re starting a brand-new product from the ground up. No legacy, no predefined stack, no “this is how we’ve always done it”.
We’re building a pragmatic LLM-based agent system - and we’re doing it properly. That means time for research, spikes, evaluation, iteration, and architectural thinking before committing to production code.

This is not just prompt tinkering.
This is about understanding how agentic systems behave - and designing them to work reliably in real-world conditions.
We’re tackling a real-world problem with a clear path to market, backed by SafeQ Cloud - and you’ll also help shape and improve the in-house tools that power the product.

What we’re experimenting with right now

• RAG pipelines
• Vector databases
• LLM vendor integrations (OpenAI, Anthropic, Google…)
• LangChain stacks
• Evaluation frameworks (e.g. DeepEval)
• Agent orchestration patterns
• MCP servers
• Observability and reliability of LLM outputs

Some of these will stay. Some won’t. That’s the point.

Your role

You won’t just “implement tickets”. You’ll:

• Help define the system architecture
• Run technical spikes and validate assumptions
• Compare approaches (framework vs custom, vendor vs open-source, etc.)
• Design how agents should behave before they are written
• Turn ambiguous ideas into technical direction
• Work closely with engineers who will later harden the solution for production

We could list the precise tech stack and final architecture…
…but that’s the fun part: we don’t have it yet.
Want to help define it?

Your toolbox

• Strong backend experience (medior+)
• Solid engineering background in Python or NET (C#)
• Genuine interest in developing agentic systems
• Comfort with uncertainty and iterative problem-solving
• Ability to think in systems, not just endpoints
• Czech and English language

Bonus if you’ve touched:

• LangChain or similar orchestration frameworks
• Vector DBs (Pinecone, Weaviate, pgvector…)
• LLM evaluation tooling
• Cloud-native architectures
• Distributed systems

What you’ll get

• Greenfield product in a stable company
• Time and space for real research (not just weekend hacks)
• Influence over architecture from day one
• A team setting - this is collaborative exploration, not a solo R&D lab