OR Scientist

TechDigital Group; Dallas (TX); On-site; USD 120,000 - 180,000; Full time

Job summary

TechDigital Group is seeking an experienced optimization scientist to advance a production-grade MIP model for a task assignment system. You will translate complex business rules into linear constraints, drive model execution via a Java API, and tune solver performance for robust, scalable results.

The role also includes data analysis in Azure ADLS and building a Streamlit dashboard to visualize optimizer outcomes.

Qualifications

• MS/PhD in OR/IE/Applied Math/CS-optimization or equivalent.
• Experience formulating real-world rules as linear constraints and objectives.
• Hands-on experience with commercial solvers and tuning Parameters.

Responsibilities

• Work on the mixed-integer programming model for a task assignment system.
• Formulate and maintain the MIP.
• Translate business rules into model constraints and objectives.
• Drive the model using the FICO Xpress Java API and inspect .lp files and solver logs.
• Tune solver performance: mip-gap, threads, seeds, determinism.
• Keep the model solver-agnostic behind an abstraction layer.
• Validate and benchmark solutions.
• Analyze Azure ADLS data to debug and surface issues in data and constraints.
• Build and maintain a Streamlit dashboard to visualize optimizer results.
• Collaborate with a Java engineer to productionize the model and with planners/ SMEs.

Skills

MIP modeling

Solver tuning

Java programming

Python programming

Streamlit dashboard

AI coding agents

Education

MS/PhD in OR/IE/Applied Math/CS-optimization

Tools

FICO Xpress

Gurobi

CPLEX

OR-Tools

Azure ADLS

Streamlit

Job description

Job Description

Work with OR&AA team members and IT partners to gather requirements, ask questions, and influence our product and business unit teams to deliver high value/impact features to our products.

Day-to-Day Responsibilities

• Works on the mathematical optimization model at the heart of a task assignment model adhering to some complex work rules.


• Formulate and maintain the MIP.


• Translate real business rules into formulation with planners.


• Drive the model through the FICO Xpress Java API, generate/inspect .lp files and solver logs.


• Tune the solver; mip-gap, mip-abs-gap, threads, random seed, determinism; reason about branch-and-bound performance and runtime.


• Keep the model solver-agnostic behind the LinearExpression / LinearConstraint / Sense / Sign abstraction so it stays testable and portable.


• Validate & benchmark solutions.


• Analyze data stored in Azure ADLS to debug the model and find issues; mine input/output datasets and solver logs to surface data-quality and infeasibility problems, derive insights, diagnose root causes, and resolve modeling bugs (e.g., bad defaults, constraint conflicts).


• Build and maintain the Streamlit dashboard (Python) that visualizes optimizer solutions - code and ship changes directly.


• Work fluently with AI coding agents (Claude / Copilot / similar) to prototype and change models, scripts, and the dashboard fast without lowering the bar on correctness or code quality.


• Partner closely as a team player with the Java engineer to productionize the model and with planners/SMEs to capture and verify rules; document assumptions and limitations.

Top 3 Mandatory Skills and Experience

1. Mixed-Integer Programming (MIP) modeling; formulating real‑world business rules as decision variables, linear constraints, and objectives; LP/IP theory; branch‑and‑bound intuition.


2. Commercial solver experience; FICO Xpress (strongly preferred) or Gurobi/CPLEX/OR-Tools/Hexaly, including solver tuning (gaps, threads, seeds, determinism) and reading solver logs /.lp files.


3. Programming to implement model & visualize model solutions; Java (the model lives in Java behind a solver abstraction); Python (incl. Streamlit for the dashboard); able to code and ship changes in both programming languages.

Also required: comfortable using AI coding agents (Claude / Copilot) for fast changes; high coding standards; a collaborative team player (open to ideas/feedback, no solo work); a fast learner.

(Education: MS/PhD in OR / IE / Applied Math / CS-optimization or equivalent.)

Nice to Have Skills and Experience

• Java 21 / Spring Boot familiarity; clean/hexagonal model design (ports & adapters); TDD (JUnit 5) for optimization models (LP-file integration tests); scheduling/assignment/routing problem experience; heuristics/metaheuristics/CP as complements to MIP; data analysis & debugging; mining datasets (incl. Azure ADLS) and solver logs to find issues and derive insights; data wrangling (CSV/Excel/SQL); Git/GitHub, MongoDB Compass, Maven, Docker; aviation / MRO domain.

Ideal Candidate

A hands‑on optimization scientist who has shipped MIP models into production, not just notebooks. They can sit with a maintenance planner, extract fuzzy operational rules, and turn them into correct, well‑scoped linear constraints then make the model solve fast enough to be useful (gap/seed/thread tuning, reformulation, branch‑and‑bound reasoning). They write maintainable, tested code behind a solver‑agnostic abstraction, validate feasibility rigorously, and communicate trade‑offs (objective design, soft vs hard constraints, default‑value handling) to both engineers and business stakeholders. Genuine FICO Xpress (or Gurobi/CPLEX) depth is the rare, critical differentiator. They hold high standards for their code, are a collaborative team player (open to ideas and feedback, no solo work), use AI coding agents (Claude / Copilot) to iterate quickly, comfortably build/change a Streamlit dashboard, and are a fast learner. Perform analysis and derive insights from data.

Team Environment and Structure

Part of the OR & Advanced Analytics (OR&AA) - TechOps/Mechanics squad. A cross‑functional team of OR scientists, software engineers, and TechOps planning SMEs, working agile with strong engineering hygiene (PR review, Sonar/JaCoCo/PIT, TechDocs) governed by the AA AI Constitution (Spec‑Driven + Test‑Driven Development). High collaboration with Maintenance operations stakeholders.

Team Fit

They define the variables, constraints, and objective behind the abstraction layer and own solution quality and solver performance, while the Java engineer owns the surrounding service, data integrations, and deployment. Together they form the core Java/Spring engineering + MIP optimization pairing that makes this a specialized OR product team rather than a standard microservice squad. They also interface directly with Maintenance Planners to validate that optimizer output beats the manual baseline.