We’re all familiar with the notion that software is everywhere, and that in some way it touches nearly every product you’ll ever own. One such product is dimensional lumber, like a 2x4 or 4x4. There are a number of steps between a tree in a forest and a piece of lumber you buy in a store. One of those is ‘edging’, the process of removing the living edge from a flat section of raw material, and producing a board of an appropriate width with straight sides.

This talk is a post-mortem of a prototype system we built for optimizing the potential value of material coming out of an edger. While the AI for optimizing produced material was an important part of our system, it wasn’t the only part of our system!

In this talk, we’ll cover: * The general problem of dimensional lumber extraction * How the client’s brand influenced which AI techniques we used to solve the problem * How AI is just a part of a larger software product, including * How we took an agile approach to AI development * How we handled estimating the cost of building the solver (and the rest of the software) * How AI integrated with the rest of the team

I’m hoping the audience takes away: * Sometimes the best technical solution is not the best overall solution * Even when AI is required for a product, it is never the whole product * AI software isn’t ‘special’ from a best-development-practices perspective