Founding Engineer — Decision Systems (m/f/d)

Muris · Berlin · first technical hire · regularly on site at the plants

Muris builds the decision system for the logistics between field and factory. Our first customer is a sugar producer. Today we deliver recommendations to their dispatchers. The goal is to run the entire chain centrally from the factory.

You are the first technical hire. There is no team you join — you help build it later. Your job is not "building features". Your job is to turn an operational mess into a decidable problem and to ship the result so that dispatchers actually use it. Everything in between is yours: data, model, system, interface.

What you do

• Formalize a real dispatching problem: what is being decided, at what granularity, under which assumptions, against which objectives.
• Choose the simplest model that carries the decision. Heuristic before solver, solver before learning.
• Define the evaluation before you build. What would a better decision have been, and how would we know.
• Ship end-to-end: from the data pipeline through the model to the interface the dispatcher looks at. A recommendation nobody understands doesn't count.
• Sit at the plant and with the dispatchers. The rules that aren't in the data — you get them there.

The first concrete problems are already on the table: night truck assignment, loading targets below the main carrier, return freight, and later dispatching without a central coordinator.

This is not for you if

• you're looking for a narrowly scoped specialist role
• you mainly integrate, ship tickets, and want to be "visible with the customer"
• you'd rather keep research questions open than put a decision into production
• you want to fit a given dataset instead of posing the problem yourself
• you're not willing to regularly sit in a sugar plant

This is for you if

• you've turned messy reality into a decision problem before (OR, dispatch, planning, pricing, allocation)
• operators have used something you built
• you're quantitatively strong and still prefer owning the whole over owning the method
• you can explain models without hiding behind the method
• you can work with incomplete data and hard constraints

Python is required. Frameworks are not. We work in English. German is a big plus — our dispatchers speak it. Polish or Ukrainian helps too; our first customer runs plants there. We're based in Berlin. The plants are not.

What you get

Real ownership: as the first technical hire you get VSOP equity — up to the low single-digit percent range depending on profile, and since we have no investors, none of it has been diluted. The percentages mean what they say. You work directly with the founders, and every week you see whether your decisions hold up at the plant. We are transparent about salary and the full package in the first call.

How we hire

A short call, then a paid work sample on a real problem from our stock: formalize it, pick the granularity, propose two model classes, argue why the simpler one goes first, define how you evaluate. Then a day with us. References, offer. Two to three weeks overall if it works on both sides.

You need an existing work permit for Germany — we can't sponsor visas yet.