Optisaur

Linear and mixed-integer optimization, in plain English

Optimization you can read.

Describe your planning problem in plain English. Optisaur writes a model you can read, solves it exactly with a proven solver, and explains the answer in plain language.

Start solving

No account and nothing to upload: the transportation example loads and solves in one click.

How it works

  1. 01Upload

    A spreadsheet, in whatever layout it is already in.

  2. 02Describe

    One sentence about what you want.

  3. 03Review

    A model you can read, every assumption spelled out.

  4. 04Solve

    A proven optimization engine finds the optimum and proves it - not the AI.

Why not Excel Solver, or a consultant?

Those are the two things a planner reaches for today. Optisaur is built to be the third.

Versus Excel Solver

Solver hands you numbers and no reasons. Optisaur shows you the model it built, the assumptions it made, which limits are binding and what a change would do - and the add-in's ceiling of two hundred decision variables does not apply.

Versus a consultant

A consultant builds the model, and the understanding leaves with them. Optisaur leaves you a model written in words you can read, change and run again yourself - in an afternoon rather than an engagement.

What Optisaur never does

The assistant never calculates.
Every number comes from a proven mathematical optimization engine, never from the AI, and every optimum carries its proof.
Nothing it writes is run as code.
It produces a structured model that deterministic code checks and compiles. There is no script to trust.
It sees the shape of your tables, not their contents.
Column names and a handful of sample values go to the assistant; the tables themselves stay with the solver.
When there is no answer, it says why.
An infeasible plan comes back with the rows that conflict and the smallest change that would resolve them.

Upcoming functionality

Today Optisaur solves linear and mixed-integer models. These are planned next; none of them is in the product yet.

  • Quadratic and Convex Nonlinear Support

    Planned

    Costs and risks that curve instead of running straight: portfolio variance, economies of scale, smooth trade-offs.

  • General Nonlinear and Nonconvex Support

    Planned

    Blending, pooling and physical processes that no straight line describes, on engines built for them.

  • Scenario-Based Stochastic Optimization

    Planned

    One plan chosen against many possible futures at once, with the adjustment each future would call for.

  • Monte Carlo / Simulation-Based Sensitivity

    Planned

    The same model solved over thousands of sampled inputs, summed up as the range of outcomes to expect.

  • Simulation Modeling

    Planned

    A process played forward step by step - queues, stock, timing - to test a plan before it is real.

See it on a real problem.

The transportation example is waiting, already solved in principle - press once to watch it happen.

Try the example