Senior AI / LLM Engineer – Agent Developer (m/f/d)

Who we are

BLOCKS is bringing big-company cloud efficiency to startups and SMEs. We guarantee 20% savings from day one by pooling cloud spend into a group-buying model. From there, our AI agent “Major Tom” acts like a virtual DevOps engineer, finding waste, optimizing infrastructure, all in an automated way to maximize cloud efficiency. And this is just the beginning, more to follow in our discussion with you.

Our founders (Oliver and Andreas) are serial entrepreneurs, have raised >$1 billion, and successfully sold multiple businesses. 

Your role

You’ll build the conversational and task-driven agents behind Major Tom, our Slack bot. You’ll build agents and MCP server tools that pull data from our data lake and data warehouse to analyze AWS billing, cloud resource usage, and related operational data. These tools will deliver useful insights and guidance to users. Over time, the agents will also be able to recommend and implement infrastructure changes to help clients optimize costs. You’ll be responsible for evaluating these agents, measuring performance, identifying and reducing hallucinations, and continuously improving the quality of their responses.

Your immediate focus: Build the first production agent, deliver it through our Slack bot, and define the architectural scaffold for our broader agent ecosystem, including how agents are created, how they talk to each other, and how they’re evaluated

As one of our founding engineers, you'll make key architectural decisions that support our growth from startup to enterprise. You'll work closely with the DevOps, AI, backend, and frontend teams to bring “Major Tom” to life.

What you bring

• Agent building: You have shipped a production LLM assistant/agent that calls tools, pulls internal data, and returns structured, accurate answers (not just a demo), using OpenAI / Claude / Gemini  style function-calling + ReAct/agent frameworks (LangChain / LangGraph / Pydantic AI)

• RAG pipelines: You have built retrieval / RAG pipelines over proprietary data: embeddings, vector DB, metadata filtering, summarization / context window management

• Agent evaluation: You have versioned and evaluated prompts and tool policies like code: regression tests, hallucination/accuracy checks, FinOps-style reasoning (“why did spend spike?“), offline eval / tracing stack (e.g. LangSmith

• Infrastructure: You have leveraged MCP-style tool server / function-calling layer to expose internal cost + usage data

• Startup execution mindset: You have shipped 0→1 products fast with high ownership, autonomy, and accountability

• Curious, pragmatic problem solver: You experiment quickly, learn deeply, and optimise for impact over theory

• Direct communicator: You value clarity and feedback as tools for speed, improvement, and alignment

• Demonstrated impact: You have impressive achievements from previous jobs and from side projects. Please share them with us.

What we offer

• Build where AI actually matters. Real AI shipped into production that changes how companies run cloud infrastructure.

• Work with founders who’ve done it before. Unicorn builders, $1B+ raised, zero politics, fast calls, radical candor. We move, not debate.

• Zero to One. The opportunity to build a product which startups will consider standard infrastructure.

• Real ownership, not theater. We challenge, you make calls. Some will be wrong. That’s fine. We fix fast and keep shipping.

• Compensation that reflects ambition. Top market cash compensation plus meaningful equity upside.

• Berlin, in person, by design. Early stage companies are built together, not on Zoom. We support relocation.

How we hire

We believe hiring should be like we build: Fast, respecting your time and ours, to the point: 

• Intro call with a founding team member (30 min) 

• Tech interview/assessment with an engineer (~60 min)

• In-person tech case study with founders and engineers (1-2 hours)

• Offer call within 24 hours with a founder (15 min)