NVIDIA's cuOpt news, custom visualizations, a preview of challenge mode & more
👋 Howdy, all. Ryan O'Neil here opening our March newsletter. The big buzz this month: NVIDIA's announcement with cuOpt.
First, there’s still a lot I’m parsing from the news and community’s reaction — much to unwrap. On the surface, it's exciting to see something like this elevate the visibility and discussion around decision science. There's a huge opportunity ahead to rethink how optimization algorithms work in the context of GPUs: What do exact optimization methods look like beyond 8 threads? 🤔
The announcement prompted a few other musings:
The future of optimization will include new and different algorithms from the ones we learned in school. GPU technology is still in active development, but it’s real. It’s revolutionized quite a few disciplines already – and now the optimization world gets to benefit, too.
The intent to make the offering open source is great. While optimization has been slower than counterparts like machine learning to adopt OSS, open source projects have shown themselves to be important in terms of adoption, prevalence, and community.
Beyond making our algorithms faster and our searches embarrassingly parallel, what else might this unlock? There are interesting opportunities for these developments to tighten the integration between ML and OR models, which I haven't seen as much discussion about. What would it look like to embed a deep learning model into a Gurobi model?
It's fantastic to see NVIDIA recognize several folks in our space (SimpleRose, Gurobi Optimization, HiGHS, AMPL Optimization Inc.). We've certainly been thinking about cuOpt in the context of Nextmv. And while most practitioners are not solving LPs that don’t fit into memory today (LinkedIn’s DuaLip is one exception in my mind), I think there are a lot of opportunities this will unlock tomorrow -- and how optimization algorithms develop. I, for one, am really excited to try out the cuOpt solvers, and to someday see what running an optimization looks like on a Grace Hopper superchip!
That’s all for now. On to the latest and greatest on the Nextmv front!
What’s new?
Render custom visualizations for your optimization model with Plotly, Chart.js, and GeoJSON
Standardize and manage custom visuals as data assets in Nextmv alongside your model. Accelerate your workflows with interactive charts, plots, and maps that automatically render alongside run details for streamlined analysis and collaboration. Learn more >>
Updated routing visuals in the Nextmv UI
The routing visualization in the Nextmv UI now uses vector-based tiles, renders more efficiently, and offers increased flexibility to include metadata and more. Learn more >>
Customer story
MatchBack Systems: Eliminating historical drayage inefficiencies for imports and exports
Optimization plays a key role in MatchBack Systems successfully minimizing costs and improving equipment utilization for its customers — but it's the speed to deployment and model iteration that takes their solution to the next level. Learn more >>
On-demand videos
Nextmv Hexaly Integration: How to run, test, and manage with DecisionOps workflows
Interested in Hexaly to solve optimization problems? This techtalk covers how to get started with Hexaly in Nextmv, use features for scenario testing, best plan auto-selection, model management, and a Q&A with Hexaly’s CEO. Watch now >>
Preview: Compare decision model outputs with a "challenge mode"
Get an early look at a feature that will allow business users, analysts, or modelers to change output data, recompute statistics, and compare to another model run. Watch now >>
Preview: Feedback and commenting workflows on decision models
Feedback on any project is inevitable. With optimization projects and decision models, we're developing more efficient commenting and feedback loops that sit alongside modeling work. Watch now >>
Community highlights
Ioannis D. Leonidas is building an open source community around optimization with over 50 optimization problems and 15 optimization algorithms | LinkedIn post
Matthew Sharp discusses the role of Python in production AI & ML work | LinkedIn post
Ryan O’Neil shares his takeaways from the solvers cluster at the INFORMS Computing Society Conference 2025 | Blog post
Events
Nextmv at INFORMS Analytics+ 2025 | April 6-8, 2025
Why puzzles are very interesting for OR consulting? (Presented by EURO Practitioners' Forum) | May 16, 2025
Jobs roundup
Looking for a job? Looking to hire? Check out our jobs board and contribute! Recently added roles:
Reddit: Senior Data Science Manager, Ads Marketplace - United States
Stitch Fix: Data Scientist | Logistics Algorithms - United States (Remote)
Uber: Senior Scientist, Pricing and Incentives - New York, NY
Audible: Senior Applied Scientist - Seattle, WA
Habitat Energy: Senior Data Scientist, Optimization - Austin, TX
Veho: Staff Data Scientist, Route Building - United States (Remote)
Open Systems: Data Scientist/Operations Research - United States (Remote)
ŌURA: Senior Data Scientist (Supply Chain) - San Francisco, CA
General Mills: Senior Data Scientist - Operations Research - Minneapolis, MN
Metropolis Technologies: Senior Applied Scientist, Pricing Optimization
Until next time!
Sign up for a free Nextmv account. Have questions? Reach out to us directly.