A coatings engineer with one obsession.

I want to end trial and error in coatings R&D

I've been working in the coatings industry since 2013. First as an apprentice at FreiLacke, then I studied chemical engineering for colors and coatings at Esslingen University, and eventually did my PhD on marine coatings. Today I work at HOBUM Oleochemicals in Hamburg, developing and testing biobased binders made from vegetable oils.

One thing I've noticed in my still rather short career is that labs everywhere run their experiments quite inefficiently. You change one factor, observe the result, change the next one, and so on. And the documentation of what came out is, let's say, not optimal.

Both together mean that a lab barely learns anything. The person who ran the experiment might, but the team doesn't really get smarter. If you're that person, you probably wrote everything down in a spreadsheet or a PowerPoint. A year from now you'll struggle to understand it yourself. And your coworkers have no idea how to read it at all. So the next project starts more or less from zero.

And I get why it has always been like that. Because proper documentation is extremely time-consuming. And setting up a system or a computer program that can learn from your past results is really difficult. If you're a small company, you don't have the expertise to do that. Well, until now.

AI, or more precisely large language models, got so good that you don't need to understand the details anymore. You only need to know what you want and be able to express that clearly. Everything else, the model takes care of in the background.

That is what all this is about. How to use and exploit the advancements in AI we currently see to build a lab that learns.

Marcel