OR Scientist

Decision Science

• Prior exposure to supply chain or a comparable operational-decision domain (inventory, transportation, planning, network design).

• A track record of work that ran at real scale — large problem instances, meaningful SKU/location/order volumes, live production traffic — rather than only academic or proof-of-concept work.

• Experience shipping something that stayed in production and that others depended on, including monitoring and iteration, not just a handed-off prototype.

• A self-initiated project or improvement you drove end-to-end without being asked — we place a lot of weight on this kind of ownership.

• A novel formulation, technique, framework, paper, patent, or open-source contribution that measurably beat the prior approach.

• Background at a company or lab doing optimization, supply chain, or algorithmic decisioning at scale (logistics, e-commerce, ride-hailing, marketplaces, manufacturing, OR-heavy tech, or a strong research group).

Full-time · Decision Science team · Reports into Algorithms/Applied Science leadership

Lyric builds an AI-native platform for supply chain decisions, where modeling, planning, and applied decision science run on a single composable architecture. The platform gives enterprise supply chain teams out-of-the-box algorithms for network optimization, inventory allocation, routing, fulfillment capacity planning, order promising, demand propagation, and scenario analysis — alongside a platform-first layer that lets business and technical users build and extend that intelligence with no-code tools or their own code.

Lyric is backed by leading enterprise software investors and works with global Fortune 500 supply chain organizations. The team is built around deep, applied expertise at the intersection of operations research, algorithms, and supply chain — and the OR Scientist / Algorithms Engineer ladder is the technical core of that team.

As an OR Scientist / Algorithms Engineer, you'll turn ambiguous, messy supply chain problems — inventory, network design, routing, planning, capacity — into correct, well-reasoned optimization models, and ship them as production-grade software. You'll work across the full arc from problem formulation through solver implementation to communicating impact, and you'll use (and help shape how the team uses) AI/LLM tooling as both a build accelerator and a product building block.

This is an entry/early-career IC role on our ladder: you'll own well-scoped pieces of a bigger problem under the guidance of senior scientists, with a clear growth path to Senior, Lead, and Principal as your scope and independence grow.

• →

  Formulate real supply chain decisions — inventory, routing, network design, planning, capacity — as LP/MIP, network flow, heuristic, or simulation models, stating and defending your assumptions.

• →

  Diagnose and improve slow or infeasible optimization models: tighten formulations, apply valid inequalities, warm starts, decomposition, or heuristics as the situation calls for.

• →

  Write clean, tested, production-grade Python (or the team's language) against solvers and modeling layers (e.g. Gurobi, CPLEX, OR-Tools via Pyomo/JuMP), and take models from prototype toward reliable, monitored production use.

• →

  Use AI/LLM and agentic tooling critically — as an accelerator you validate line-by-line, and, where relevant, as a component you help design, evaluate, and guardrail inside an algorithmic pipeline.

• →

  Connect every model back to the business decision it serves — service levels, lead times, plan stability — and recognize when an “optimal” plan is operationally useless.

• →

  Explain your formulation, trade-offs, and results clearly to both technical and non-technical audiences, and take feedback as input rather than verdict.

• Hands-on experience formulating and solving real optimization problems — LP/MIP, heuristics, network flows, or simulation — beyond coursework alone.

• Professional software engineering practice in Python (or a comparable language): version control, testing, and code review, not just notebooks.

• Exposure to at least one commercial or open-source solver via a modeling layer (e.g. Gurobi, CPLEX, OR-Tools, Pyomo, JuMP) — or clear evidence you pick this up fast.

• A critical, hands-on posture toward AI/LLM coding tools: comfortable using them, and equally comfortable explaining where they've gone wrong.