About

I'm Nate Cheever. I build things that predict things.

I've done data work since 2013: Excel and macros first, then R, then Python after a data science bootcamp in New York in 2015. NLP has been my favorite part of it ever since. I'm almost entirely self-taught: no CS degree, no grad school, which mostly means I learned the fundamentals the hard way, by needing them.

In 2017 a friend from college texted to ask if I knew much about machine learning. His company needed a churn model, and, though I didn't know it at the time, I was one of three contenders: me, and two much larger data science firms, all building against the same dataset. I won. That became eight years at AdvancedMD, where I built their automated data pipelines, their analytics database, and an internal dashboard tool called Beat that the company still runs on today.

Earlier this year, AdvancedMD brought in an enterprise vendor to see how their platform compared to my existing model. Mine beat it by 32 points of recall. I used that head-to-head result to make the case for going independent, and AdvancedMD stayed on, not as my employer, but as my first client. That's Pydealist.

What I do now

The same work, aimed at organizations with a mission instead of a shareholder. Donor retention modeling, segmentation, and the dashboards that make the numbers usable, for nonprofits, foundations, and the people who run them.

Every engagement is with me directly. I don't hand you off to an account manager, and I don't build a model I can't explain in plain language.

Outside of work

I live in Stillwater, Oklahoma, with my wife and four kids, on about an acre.

I read more than is probably efficient.

I keep a physical zettelkasten of notecards on philosophy. Over a thousand cards now.

I used to run a physical newspaper for this town.

I run a local newsletter here now, Stillwater Lokal Life. I love local media.

If you want the numbers, book 20 minutes. If you want to know how I think about this work, that's most of what's on the blog.