OR Scientist

What matters most

• Production-Grade Optimization Modeling

  The role requires moving beyond academic prototypes to robust, scalable models. You must demonstrate proficiency in formulating LP/MIP, network flow, or simulation models and implementing them using solvers like Gurobi or OR-Tools within a production-grade Python codebase.

• Bridging Algorithmic Logic and Business Utility

  You must connect complex optimization results to real-world business outcomes such as service levels and plan stability. A key priority is identifying when a mathematically optimal solution is operationally impractical and communicating these trade-offs to stakeholders.

• Critical Integration of AI/LLM Tooling

  You are expected to use AI/LLM tools both as a development accelerator and as a functional component of the product. This requires a critical, hands-on posture where you validate AI-generated code and design guardrails for agentic pipelines.

Likely challenges to think through

• Balancing Model Performance and Complexity

  Real-world supply chain problems often lead to slow or infeasible models. You will face the challenge of diagnosing performance bottlenecks and applying advanced techniques like decomposition, warm starts, or valid inequalities to ensure models remain performant at scale.

• Navigating Ambiguous Problem Spaces

  Supply chain requirements are often messy and ill-defined. You must prepare to translate vague business needs into structured mathematical formulations while defending your assumptions and iterating based on feedback from senior team members.

Topics to prepare

• Optimization Model Performance Tuning

  The role requires diagnosing and improving slow or infeasible models. Exercise: Take a complex MIP formulation and practice applying techniques like valid inequalities, warm starts, and decomposition to reduce solve times. This is directly relevant to the requirement of moving models from prototype to production-grade performance.

• Supply Chain Decision Modeling

  The core of the role involves formulating inventory, routing, and network design problems. Exercise: Review standard formulations for facility location or vehicle routing problems and document the assumptions required to make them operationally useful. This prepares you to defend your modeling choices during technical discussions.

• AI/LLM Integration in Algorithmic Pipelines

  Lyric uses AI/LLM tools as both a development accelerator and a product component. Exercise: Research strategies for guardrailing LLM outputs within an optimization pipeline. This addresses the requirement to use AI tools critically and design them as reliable components of a larger system.

Skills to revise

• Production-Grade Python Engineering

  The role demands software engineering proficiency beyond notebooks. Exercise: Refactor a previous optimization script into a modular, tested Python package using version control (Git) and unit tests (pytest). This is essential for the requirement of shipping reliable, monitored production code.

• Solver Modeling Layers

  You must be proficient with solvers like Gurobi, CPLEX, or OR-Tools. Exercise: Implement a supply chain optimization problem using a modeling layer like Pyomo or OR-Tools. Focus on writing clean, readable code that separates the data, the model formulation, and the solver configuration.

• Prepare a production-grade optimization project walkthrough

  Select a past project where you moved an optimization model from prototype to production. Be ready to explain your choice of solver (e.g., Gurobi, OR-Tools), how you handled model performance (e.g., warm starts, decomposition), and how you ensured the code was maintainable, tested, and version-controlled.

• Demonstrate critical AI/LLM tooling proficiency

  Prepare examples of how you use AI coding assistants to accelerate development while maintaining rigorous validation. Be ready to discuss specific instances where you identified and corrected errors generated by AI tools, demonstrating your ability to act as a guardrail for automated outputs.

• Articulate business-driven optimization trade-offs

  Practice explaining a complex technical model to a non-technical stakeholder. Focus on how you translated mathematical objectives (like LP/MIP constraints) into business outcomes such as service levels, lead times, or plan stability, and how you handled scenarios where the 'optimal' solution was operationally impractical.

• Formulate thoughtful questions for the team

  Prepare questions regarding the team's current approach to integrating agentic tooling into their supply chain platform, how they balance model accuracy with computational speed, and how they manage the transition from research-based modeling to production-grade software deployment.