Whenever I start a new optimization project on my own, there’s a honeymoon phase. I get to explore a new problem and read literature, find (or generate!) input data, choose a solver (or solvers!), and formulate a number of alternative models. I get to do fun science!
Once I get through that initial development, the hard reality of choosing and deploying a production model inevitably rears its head. How do I test and compare different approaches to the problem? How do I share these results with clients? Where should I actually put the model and how does someone actually use it?
Closing the gap between idea and impact for decision models is why we’re building Nextmv. Ideas can come easily: ☝️"Adjusting service times for appointments will increase customer satisfaction!" Implementing them less so: 🤔"How do I do this without foiling worker experience?"
Whether you’re new(er) to optimization or a Level 12 Logician of the Second Order Cone 😉, the only place to start is the beginning. And the beginning often starts with you. Yes, you, a solo developer with an idea that can Newton step into something incredibly wonderful for a larger team to collaborate on. Which is why I want to highlight some recent updates we've made.
First, we've made several new decision apps available on our marketplace. Select the one you want to run, clone and customize it, or view the code in GitHub (more on that next). They’re all available to try for free.
Second, we've created an open source GitHub repository containing the model code for each of these marketplace apps. The code is there for you to fork, customize, and run with your Nextmv account (or even run somewhere else).
Anyone can open PRs against the models we've created. If you have a model you think is cool, you can request to add it to this repo to share and develop with the community. There is also a path for us to work with contributors to make models in the repository official decision apps in the Nextmv marketplace.
Third, we're designing an individual plan. We've focused a lot of our plans around teams and collaboration for custom decision modeling. But we know noteworthy projects can start small with individual decision algorithm developers tinkering with a model.
Whether you're a solo researcher, hobbyist, or a developer/decision scientist noodling on a new concept, we're designing a new plan that supports you to quickly deploy, test, and manage custom decision models. If you'd like to get early access and provide feedback, you can sign up here. Just look for the hard-to-miss orange box!
There's certainly more ground to cover with the work I've shared here today. But we think it's a step in the right direction — it's what we would have reached for when starting out on many of the optimization projects in our past lives.
Feedback, as always, is wanted and welcomed. Share your thoughts in our forum or drop us a line with a direct message.
May your solutions be ever improving 🖖