Applied Operations Research Engineer

Hi, I'm Erik, CTO at Circonomit. We build the decision platform industrial companies use to plan their production. The toughest challenges we're facing: A model that answers for one site has to keep answering across twelve, over more periods and harder constraints. And a hundred customers have to solve at once without noticing each otherOur engineers, and eventually our customers, should build and extend models without you in the room. That means deciding where the abstraction belongs and what the modeling vocabulary has to coverWhen the honest answer is "impossible", the planner still needs something usable: which rules collide, and what it would cost to bend oneI need someone to own the engine our models run on: decide how it gets built, build it, keep it running. Our models decide what gets produced when, on which machine, and at what stock level. The engine sits between the mathematics that makes the answer correct and the product that has to make it usable by people who are not mathematicians. We came out of RWTH research, raised €2.8M with Vorwerk Ventures, and have customers running on our models today. Small team, Cologne office. Requirements You have modeled and shipped combinatorial optimization in industry (MILP, CP, or both), with models that ran on messy data and real usersRecent, substantial hands-on experience writing and operating production Python: tests, types, review, measurement before optimization. You know how numerics go wrong where a model meets a solver: scaling, tolerances, integralityYou know where solvers reach their limits, CP-SAT and Gurobi included, and can say which technique bought you what: warm starts, rolling horizon, relax-and-fix, aggregation, or a heuristicExperience running optimization workloads in production, not only in notebooks: cancellation, timeouts, and parallel solves includedYou can name the modeling or solver decision that will decide whether something holds up at scale, before it is builtThe ability to make technical decisions independently and explain the trade-offsYou want to understand the customer's problem and their data, not only the mathematicsGerman at C1 level or above, and fluent English. Team communication is in German; code and documentation are in EnglishExisting work visa for GermanyYou should enjoy turning ambiguous customer problems into models that hold up, rather than waiting for a specification. Priorities change as we learn from customers, and we make scope and trade-offs explicit together. Benefits Ownership. The engine every customer model runs on: modeling language, compiler, solver integrationArchitectural influence. You'll own how modeling and solving evolve here, and make those trade-offs directly with the teamLeverage. You do not build one model for one customer. You build what every model here is built withReal-world impact. Customers plan their production on our models todayDirect collaboration. Small team, no org chart. We work directly together. Weekly feedback, both waysCompensation. Competitive salary plus a VSOP package reflecting your contribution and development in the roleHybrid work. We work from our Cologne office, with one to two home-office days per week. You don't need to live in Cologne, but you do need to be able to work from the office on the remaining daysYour setup and everyday extras. Hardware of your choice · AI tooling budget · sports membership · Deutschland-TicketProcess: A 20-minute call, a technical conversation with me and one of our engineers, a three-hour challenge with a 45-minute walkthrough, then a conversation with the team. Two to three weeks overall, with feedback within days.