MaintainX - Senior Applied Scientist, Scheduling and Optimization

• Own and evolve the Python optimization service that powers the Scheduling Agent, modeling, solving, and iterating on the constraint formulation as new use cases emerge. • Design and implement increasingly sophisticated scheduling capabilities: trade and crew constraints, irregular capacity patterns, production downtime windows, multi-site considerations, and reactive re-scheduling. • Build and maintain API routes and tools that expose the solver to GenAI agent workflows (tool calling, structured input/output). • Partner with PM and design to translate messy real-world scheduling problems into solver constraints, and push back when "optimal" isn't what users actually want. • Iterate the solver with real users via design partnerships and pilot deployments. Take feedback from human schedulers seriously and reflect it back into the model. • Contribute to the surrounding Python service: performance, observability, testing, and reliability of the optimization runtime. • Help shape how scheduling intelligence integrates with the broader MaintainX product over time, including learning from execution data to improve solver inputs. • 5+ years of professional software engineering experience, with significant time spent on optimization, constraint programming, or operations research problems shipped to real users. • Strong fluency with CP-SAT and at least one other optimization paradigm (MILP via Gurobi/CPLEX/HiGHS, metaheuristics, or similar). You've hit the limits of one approach and made informed choices about when to use which. • Solid Python service engineering: APIs, async, testing, profiling, observability. You can own a production service end-to-end. • Academic grounding in Operations Research, Industrial Engineering, Computer Science, or a related quantitative field, at minimum a strong undergraduate foundation; advanced degrees are common in this space but not required. • Track record of iterating optimization systems with real users, you've felt what happens when a human rejects the "optimal" answer and you've redesigned the model in response. • Product mindset and delivery orientation, you ship, you measure, you iterate. You think about the user outcome, not just the objective function. • Comfort with ambiguity. You can co-design the constraint data model with the team rather than waiting for a clean spec. • Familiarity with GenAI tooling (LLM tool calling, structured output, prompt design for constrained generation) is expected.