Principal Operations Research Scientist

Quvia•Washington, DC

1d•Hybrid

About The Position

Quvia is deploying Grid, our AI-powered platform that dynamically routes network traffic to optimize connectivity and performance, with enterprise customers in the U.S. We're looking for the engineer or scientist who makes it perform. The core problem is getting finite satellite bandwidth to the right users at the right time, under contested and fast-changing conditions. Capacity shifts as satellites move, demand is bursty and mission-driven, and the answer has to be both good and fast. If you've built scheduling, network flow, or supply-chain optimization systems at scale, you already know the math. The domain is the only new part. You'll be embedded with enterprise customer as the senior technical owner of Grid's optimization layer. This is hands-on work, not advisory. You'll formulate the models, write the code, choose the solvers, and profile the results yourself. You'll also translate in both directions: turning mission constraints into problems Grid can solve, and explaining what the models can and can't do in terms program leads can act on.

Requirements

U.S. Citizenship - Non-negotiable for this customer engagement.
8+ years of professional experience developing and deploying optimization or operations research systems in production . Seniority and judgment are an explicit part of the bar; this is not a role we will fill with a strong junior.
Advanced degree (MS or PhD) in Operations Research or a closely related quantitative field. Computer Science and Electrical Engineering backgrounds are equally welcome where the optimization depth is there. Equivalent industry experience will be considered in place of the degree.
Deep, applied command of optimization techniques: linear and mixed-integer programming, network flow, constraint programming, column generation or other decomposition methods, metaheuristics, and large-scale scheduling — with the practical experience to know which to reach for and when.
Strong software engineering ability. You will write and ship the models yourself.
Production-quality code in a compiled or high-performance language (Java, C++, or similar) alongside Python for modeling and analysis.
Hands-on experience with commercial and open-source solvers — Gurobi, CPLEX, OR-Tools, HiGHS, or similar — including modeling APIs, parameter tuning, and callbacks.
Track record of working directly with external customers or stakeholders on complex technical engagements, with the communication skills to hold a technical conversation and a program conversation on the same day.

Nice To Haves

AI/ML experience , particularly where learned components sit alongside or inside optimization systems: demand forecasting feeding an allocation model, learned heuristics or branching rules, or reinforcement learning for sequential decisions.
Background in supply-chain, logistics, or delivery network optimization — the kind of work done at Amazon, DoorDash, Walmart Labs, or comparable operations at scale.
Queueing theory and congestion modeling — characterizing contention, latency, and service levels when demand exceeds available capacity.
Discrete-event and Monte Carlo simulation , and simulation-optimization where an analytical model alone won't capture the system's behavior.
Stochastic, robust, and multi-stage optimization — recourse models, scenario generation, chance constraints, and designing for uncertainty rather than around it.
Dynamic programming, Markov decision processes, and approximate DP for sequential allocation and control problems.
Multi-objective optimization and trade-off analysis , including how to present a Pareto frontier to a decision-maker who has to choose a point on it.
Graph and network algorithms — shortest path, max-flow/min-cut, matching, and network design at scale.
Forecasting, time-series, and statistical modeling feeding operational models, plus design of experiments and sensitivity analysis to test how far a solution holds.
Algebraic modeling languages — Pyomo, JuMP, AMPL, or GAMS — for rapid formulation and prototyping.
High-performance and parallel computing applied to optimization: distributed solves, parallel heuristics, and profiling large model runs.

Responsibilities

Formulate and build optimization models: Take the customer's operational problems — bandwidth allocation, routing, scheduling, capacity planning, contention resolution — and turn them into rigorous mathematical formulations, then implement them in production code.
Develop and extend Grid's optimization capability: Modify and configure Grid directly against the engagement's requirements, and write new solvers, heuristics, and decomposition schemes where the existing capability doesn't reach.
Own solver and algorithm selection: Decide where exact methods (MILP, network flow, constraint programming) are the right tool and where the problem demands heuristics, metaheuristics, decomposition, or a hybrid — and be able to defend the choice.
Tune for hard runtime constraints: Live deployments impose real time budgets. Profile, reformulate, tighten relaxations, warm-start, and cut where needed to get quality solutions inside the window.
Validate rigorously: Build the test harnesses, benchmark instances, and solution-quality measures that show a model is behaving correctly, including against edge cases the customer will eventually hit.
Instrument and diagnose live systems: Debug degraded solution quality, infeasibility, and performance regressions in deployed environments where you cannot simply re-run offline.
Document formulations, assumptions, and model behavior to a standard that holds up under government program review.
Feed the roadmap: Bring field learnings back to Quvia's OR and engineering teams and identify where new optimization capability is needed in the product.
Support the full engagement lifecycle: initial configuration, pilots, evaluation, and steady-state operation.