The Nextmv AMPL integration: Accelerating decision model development with DecisionOps

Do you model optimization problems with AMPL? The Nextmv AMPL integration accelerates how you run, test, and deploy AMPL decision models using DecisionOps tooling and workflows.

The Nextmv decision modeling and solving ecosystem keeps expanding with the Nextmv AMPL integration! This new feature helps AMPL users accelerate decision model development with DecisionOps tooling and infrastructure. 

See a speedrun that covers the basics of the integration and read on for more details 👇👇👇. 

AMPL is a well-established decision modeling solution that features a large set of integrations for databases and collaboration tools, along with support for multiple solvers (e.g., Gurobi, CBC, HiGHS) and problem types. (In my time at Nextmv, the iconic AMPL book — the guide for modelers at all levels of experience by Robert Fourer — has made many appearances on video calls with our CTO Ryan O’Neil. The hardcover version, no less!)

Many in the optimization community get their start with AMPL as students and take it with them into their research or industry careers. It’s a great option that taps into the vast ecosystem of optimization tools, giving users the freedom to explore and invest in the approaches that work for the problem they’re solving. We’re excited to bring DecisionOps support to AMPL users to accelerate ways to test, deploy, and operate AMPL decision models. 

Deploy AMPL models to production-ready infrastructure in minutes. Create a scalable decision microservice complete with a set of unique API endpoints that reads from databases, aggregates decision data, submits runs, receives results, and sends plans to your operational systems. 

Set of unique API endpoints for AMPL decision app.

Perform historical and online experiments on AMPL models. Set up batch, scenario, and acceptance tests that use historical data to check model changes such as adding or removing a constraint, changing an objective function, or swapping a solver in the AMPL universe. Use existing input sets or create them from previous model runs for future historical testing (!) and share results with stakeholders. Production or online tests like shadow and switchback (A/B) let you experiment under real-world conditions in a measured way.

Examples of an acceptance test result and switchback test setup with plan summary in Nextmv console.

Manage models and collaborate via UI with stakeholders. Manage model versions and configurations like you would any other software with Git-based and CI/CD workflows. Share run history, run results, model experiments, and more with teammates and stakeholders to make collaborating on model development and confidently rolling models to production more seamless. 

To get started with the Nextmv AMPL integration, all you need is a free Nextmv account and a trial. From there, you can get up and running with a few commands:

  1. Download and install the Nextmv CLI.
  2. Clone a Nextmv AMPL app example (with our CLI: nextmv community clone -a knapsack-ampl). These community apps are also available in GitHub.
  3. Push your model code to your decision app to start making runs.

When you are ready for industry-grade use, reach out to AMPL to get a license and access commercial solvers and unlimited problem sizes. Check out the Nextmv AMPL integration documentation to learn more. 

We’re excited to make this integration available to the community and look forward to folks trying it out. We welcome feedback in our forum, or you can always reach out to us directly

May your solutions be ever improving 🖖

Video by:
No items found.