Katherine Anderson: Hi, I’m Katherine from Freya Systems. I’d like to discuss the difference in approaches we have for programming models in industry versus academia.
The main distinction is that in industry we’re focused on productionizing models, usually in a given time and budget. And in academia, we would be more focused on researching complex algorithms, to find the most innovative and elegant solution.
We aren’t rewarded in industry for how complex a model is because the value we provide goes beyond the black box of input data here, output prediction there. It is better for us to find an analytical solution that meets the need, while being easier to implement and typically more robust. We would much rather have a solution that is easier to explain the inner workings of, so that everyone involved can gain a greater understanding of the problem.
In academia, one might not approach the problem in a way that makes sense for longterm use, where there will be code maintainers and new projects using similar functions, as the goal may be more research focused, perhaps writing a paper or developing new types of models.
When I’m working on a project, I still like to start out in the quick and dirty exploratory stage where I’m learning all I can about the problem and trying new techniques but I then refactor into clean, readable code for code reviews and better version control. With this approach, I can more easily and efficiently collaborate with others who are involved in productionizing my model and establish lasting solutions that are beneficial to the end user.
I’m Katherine from Freya Systems. Thanks for watching.
