Consulting & Collaboration

Optimization-Driven Decision Support

Many organizations rely on spreadsheets to manage staffing, scheduling, capacity, demand, and resource-allocation decisions. These spreadsheets often contain valuable operational knowledge, but they do not automatically identify the best decision or verify whether all constraints have been satisfied.

My work focuses on converting these informal planning processes into rigorous mathematical decision models. This includes identifying decision variables, operational constraints, objectives, uncertainty, and tradeoffs, and then developing computational tools that generate implementable recommendations.

The value is not simply automating a spreadsheet. It is designing a mathematically sound optimization model that captures the real decision structure, recognizes feasibility and tradeoffs, and produces solutions that can be evaluated, explained, and improved.

Featured Prototype: Spreadsheet-to-Decision Scheduling

Functional Proof of Concept

Role-Aware Staffing and Scheduling Optimization

I developed a complete prototype that transforms a semi-structured staffing workbook into a validated, optimization-ready dataset and then generates a recommended schedule.

The prototype demonstrates the complete process from raw operational data to an optimized decision.

Existing Spreadsheet Employee data, availability, preferences, roles, and shift requirements

Data Normalization Detection, transformation, cleaning, and validation of company-specific formats

Optimization Model Mixed-integer formulation of staffing requirements, constraints, and operational priorities

Decision Output Recommended schedule, shortages, workloads, preference use, and management summaries

Current Model Capabilities

Role-specific coverage requirements for RN, LPN, CNA, and other employee classes
Employee availability and infeasible-assignment restrictions
Soft penalties for undesirable but allowable assignments
Maximum workload and no-consecutive-shift constraints
Explicit role-specific shortage variables
Optional labor-cost and workload-fairness objectives
Automated schedule, shift-coverage, and employee-workload reports

Technical Components

Canonical data schemas for employees, shifts, availability, and staffing requirements
Company-specific spreadsheet adapters and column-mapping logic
Data validation before optimization
Mixed-integer optimization models developed in Pyomo
Open-source mathematical programming solvers
Scenario analysis and model diagnostics
Extensible architecture for different organizations and spreadsheet formats

What Requires Optimization Expertise?

Building a decision-support tool requires more than reading an Excel file and writing software. The central technical task is formulating the organization’s operational problem correctly.

Determining which quantities are decisions and which are fixed inputs
Translating operational policies into mathematically precise constraints
Distinguishing hard requirements from soft preferences
Designing penalty structures that reflect real operational priorities
Identifying infeasibility and explaining why requirements cannot all be satisfied
Balancing competing objectives such as staffing coverage, cost, fairness, service quality, and risk
Selecting algorithms and formulations that remain scalable as the problem grows
Validating that the computational output is meaningful, implementable, and consistent with the intended model

These are mathematical-modeling and algorithm-design questions, and they are where advanced training in optimization and operations research provides substantial value.

Areas of Expertise

Optimization and Operations Research

Linear and mixed-integer optimization
Convex and conic optimization
Distributed and network optimization
Staffing and scheduling
Resource allocation
Multi-stage and sequential decision-making
Optimization under uncertainty

Machine Learning and Applied AI

Predictive modeling and forecasting
Machine-learning-assisted decision systems
Reinforcement learning
Graph-based learning and network models
AI-assisted data interpretation
Forecasting-to-operations pipelines

Data and Software Development

Spreadsheet-to-dataset transformation
Data validation and schema design
Python, Pyomo, PyTorch, and scientific computing
Optimization software prototyping
Scenario-analysis tools
Reproducible analytical pipelines

Application Domains

Power and energy systems
Staffing and workforce planning
Maintenance optimization
Healthcare analytics
Supply-chain and network planning
Engineering and operational analytics

Collaboration Process

Problem discovery: review the current planning process, spreadsheet, and operational objectives.
Data and model assessment: identify available data, decision variables, constraints, performance measures, and missing information.
Mathematical formulation: develop an optimization or decision model that reflects the organization’s operating rules.
Prototype development: build a working computational pipeline using sample or historical data.
Validation: compare recommendations with current practice, test scenarios, and refine assumptions.
Deployment planning: determine whether the prototype should remain a decision analysis tool or be extended into an operational application.

Potential Engagements

Assessment of an existing spreadsheet-based decision process
Fixed-scope proof-of-concept optimization prototype
Staffing, scheduling, routing, allocation, or capacity planning model
Forecasting-to-operations decision pipeline
Scenario-analysis and planning-under-uncertainty tool
Mathematical review of an existing analytical or AI system
University-industry research collaboration
Grant-supported applied research partnership

Background

I am an Assistant Professor of Mathematics and Data Science at Illinois State University. I hold PhDs in both mathematics and electrical engineering.

My research lies at the intersection of optimization, machine learning, mathematical modeling, and data-driven decision-making. My work spans theoretical foundations, algorithm development, optimization software, and applications in power systems, operations research, healthcare, maintenance, and engineering analytics.

This combination of mathematical depth, engineering background, and implementation experience allows me to work across the full project lifecycle—from problem formulation and algorithm design to prototype development and computational validation.

Discuss a Pilot Project

I am interested in small, focused pilot projects involving spreadsheet-based planning, staffing, scheduling, forecasting, resource allocation, or operational decision support.

A pilot can begin with one existing spreadsheet, one clearly defined operational problem, and a limited-scope prototype.

Contact Me

Contact

If your organization has an operational planning problem and would like to explore whether optimization or applied AI can help, please feel free to reach out.

Email: mkarim3@ilstu.edu

LinkedIn: linkedin.com/in/mehdi-karimi-91324b115

GitHub: github.com/mehdi-karimi-math