My LP Solver

LP Solver: Solving Linear Programming Problems with AI

Linear programming is one of the most powerful tools in operations research and decision-making. It is used daily by companies to optimize production, minimize costs, allocate resources, or plan logistics, yet it has long remained the domain of specialists who know how to formulate mathematical models.

This application changes that.

What does it do?

LP Solver is an AI-powered web application that allows anyone, with or without a mathematical background, to solve linear programming problems in two ways:

Natural language: simply describe your problem in plain English, as you would explain it to a colleague. The AI understands your intent and automatically builds the mathematical model. LP format: if you are familiar with linear programming notation, you can paste a structured LP block directly and let the solver handle the rest.

In both cases, the application extracts the optimization model, solves it instantly, and displays the results clearly: the optimal objective value and the value of each decision variable.

A concrete example

Imagine a factory producing two products, A and B. Each unit of A generates €5 of profit and requires 2 hours of machine time and 1 hour of labor. Each unit of B generates €4 and requires 1 hour of machine time and 2 hours of labor. The factory has 100 machine hours and 80 labor hours available per week. What production plan maximizes profit? Just type this situation in plain English and the application returns the answer in seconds: produce 40 units of A and 20 units of B for a maximum weekly profit of €280.

How does it work?

The application is built on a lightweight agentic architecture connecting two components:

LLM Understands the problem and extracts the mathematical model
PuLP / CBCSolves the linear program with a proven open-source solver
The user input, whether free text or structured LP notation, is sent to the large language model, which returns a structured JSON representation of the optimization problem.
This model is then passed to PuLP, which builds and solves it using the CBC solver. Results are displayed immediately in the interface.

Why does it matter?

This project demonstrates how large language models can act as intelligent bridges between human intent and formal mathematical tools making operations research accessible to a much wider audience without sacrificing rigor or accuracy. It is also a working proof of concept for agentic AI workflows: the LLM does not just generate text, it produces structured, validated output that drives a downstream computational process.

Try it

The application is freely accessible at the link below. No installation, no account required, just describe your optimization problem and let the solver do the work.

👉 lp-solver-souissi.streamlit.app

⚠️ Disclaimer: This is an experimental application. The AI-generated model may occasionally misinterpret complex problem descriptions. Use results with caution.

Source code available on GitHub under CC BY-NC 4.0 license (non-commercial use only).