NumFOCUS signs agreement with LANL to develop and support an AI toolbox for JuMP
announcements ·Author: Miles Lubin, Julian Hall, and Changhyun Kwon
The JuMP Steering Committee is pleased to announce that we, through NumFOCUS, have signed an agreement with Los Alamos National Laboratory (LANL) to develop an AI toolbox for JuMP.
The main deliverable is a new package, MathOptAI.jl, which is a package for
embedding trained machine learning predictors into a JuMP model. MathOptAI is
inspired by packages such as
gurobi-machinelearning
and OMLT. (We will update this post with
a link to MathOptAI once it is public.)
A second deliverable is to develop a toolbox for creating, deploying, managing, and analyzing computational experiments related to JuMP on LANL’s HPC clusters. The goal of this task is to simplify experiments such as Carleton Coffrin’s AC-OPF benchmark.
As a third deliverable, we will provide general support for LANL staff and their project partners on JuMP and Julia for AI.
The agreement runs until September 2026, and the person being funded is Dr. Oscar Dowson, a core contributor to JuMP, and also a member of the Steering Committee.
(Per our conflict of interest policy, Oscar, who is funded by the grant, and Carleton, who works at LANL, recused themselves from Steering Committee votes on this agreement.)