Welcome to Gurobi AI Modeling!

gurobi-ai-modeling aims to familiarize users with using generative AI for mathematical modeling. This repository contains documentation that features the following:

• Help users understand the capabilities and limitations of the state-of-the-art generative AI tools in the context of optimization.

• Propose best practices for using generative AI to build and solve optimization models and describe common pitfalls.

• Share examples of prompts and expected outcomes.

• Describe tools that can help create and solve models.

Contents

Introduction

• About this repository
  • How to utilize this repository
  • Structure
  • What to take away from this
• Getting Started
  • Using ChatGPT as a translator for mathematics
• Use cases
  • Specialties

LLM Prompting

• Writing your first prompt
• Example prompts
  • Classic OR Examples
    • Facility
    • Diet
    • Sudoku
    • Workforce
    • Portfolio Optimization
  • DevOps Engineer
    • Scheduling deployments
    • Assigning workloads
    • Incident Response Planning
    • Testing strategy optimization
  • Backend Engineer
    • Ad Selection
    • A/B Testing
    • Backend API Routing
    • E-Commerce Pricing
  • Data Engineer
    • Apache Kafka Partition assignments
    • Snowflake Optimizing clustering keys
• A deeper look
  • Prompt engineering
  • Answer engineering
    • Interpreting the question
    • Generating a mathematical representation
    • Generating a model
    • Interpreting the results
• Custom GPTs
  • Gurobi AI Modeling Assistant
    • Advantages
    • Prompt Template
  • Gurobi AI Modeling Prompt Engineer
  • Gurobot
• Tips and pitfalls
  • Write problems as unambiguous as possible
  • Unexpected prompts can lead to unexpected behavior
  • Should variables be considered divisible or not?
  • Avoid unnecessary words or statements
  • Supply all necessary (dummy) data
  • Technical Issues
    • The LLM cannot install the wheel or cannot read attached data files
    • LLM is generating code but not executing it
  • Correct usage of gurobipy
    • Avoid usage of recent gurobipy API’s
    • Too much gurobipy output
  • Other modeling Pitfalls
    • Typos
    • Mixing data types
    • Too much inline data
    • Data preprocessing
  • Examples of LLM Reasoning limitations
    • I blindly assume that supply and demand are equal
    • I don’t know how to store inventory for later use
    • I have difficulty thinking in 3D space
    • I don’t understand how hosting applications works

Learning Optimization

• Comparing to Machine Learning
  • Machine Learning is excellent at predictive modeling
  • Mathematical Optimization is a prescriptive approach for decision-making
  • To sum up, there are big differences
• Modeling
  • The Problem Statement
  • Components of an Optimization Model
  • Translation of Problem Statement into Mathematical Formulation: Two Options
  • Review, Validate and Iterate

Evaluating and troubleshooting models

• Testing
  • Ensure that the objective value is plausible
  • Visualize the solution
  • Verify that the solution satisfies your constraints
  • Other specific edge cases that indicate validity

More Educational resources

• Learn Mathematical Optimization
  • Hands-on Optimization Modeling
  • Blogs and Podcasts
  • Online Courses and Guided Training
  • Gamify Optimization!

Help and Feedback

Press / or click to search this documentation	Gurobi website	Gurobi community forum
Go to other Gurobi documentation	Github Repository	Take our brief survey!