Kira McLean (4)
This is also an area where Clojure really shines. It’s not the easiest to get started with, but once you’re up and running it is miles easier to test, version, deploy, package, scale, and repeatedly run your pipelines or models reliably than anything else I’ve tried so far. Paying the startup cost is well worth it if your goal is a robust production system that works.
First keynote was amazing. I love hearing from other successful women in my field. Also super validating to hear from a real industry veteran that the time has come to bring robust engineering best practices to data science. A lot of companies do this well but there is a long way to go.
In software we’ve been working on “best practices” and processes for a couple of decades. So much of it is generic enough to be useful for data science too. We can help!
Learning about PyMC makes me want to become a statistician.. super interesting way to think about data, but so much goes into building a good model! So many rabbit holes.
Bayesian modelling is clearly super powerful though and seems to offer some answers to some of the most intractable problems with black-box ML. A reliable model with known and understandable inputs is invaluable for certain use cases.
