The problem: setting up an ML-driven experiment takes steps most scientists skip
Machine learning can suggest the next formula from past results, but setting up targets, constraints and the model run by hand takes time and know-how.
Most formulators know their project data could tell them what to try next. In practice, setting up a predictive model means defining targets, ranges and constraints in a tool built for data scientists, so the model often never gets run and the next experiment is chosen by intuition.
What this demo shows
- Start from a weathering project with recipes and measurements already in Uncountable.
- Ask Bodie for the next coating with tensile strength above 13 and Delta E at or below 0.7.
- Bodie builds the constraints and runs the Suggest Experiments job in the background, then sends a notification.
- Review the suggested recipe with effect sizes, surface plots and visualizations, and import it into the project.
Why it matters
Each round of lab work costs time and materials. A suggestion grounded in the project's own recipes and results gives the team a better starting point than a guess, and the effect sizes show which ingredients are worth adjusting.
Asking in plain language lowers the barrier to using predictive tools at all. The scientist states the targets, Bodie sets up and runs the job, and the suggested recipe lands in the project ready to test.
See it with your own data
This video is part of DOE and Predictive Tools. To see how it would work with your own formulations and records, request a demo. Our team replies within 30 minutes during business hours.
Read the transcript
Now, in Uncountable, you can ask Bodie to suggest your next experiment using our suggest experiment machine learning models. As you can see here, here I have some experiments in this Suncil weathering program project with corresponding recipes and measurements. And here I'm asking Bodie to suggest the next Suncil coating to run with tensile strength greater than 13 and delta E smaller than or equal to 0.7 while keeping [unclear] below 12.
As you can see here, Bodie created a constraint aspect based on my instruction and started the suggest formulation job in the background. All this done automatically without me having to intervene at all. Once Bodie is done with the job, it will send you a notification to tell you that it has finished the background task and provide me with details of the results and suggested recipes for me to review.
As you can see here, Bodie has already suggested me the next coating to try. I can simply ask Bodie to take me to the result page. As you can see here, Bodie has created for me the suggested coating. I can read the results here with the suggested recipe that I can easily import into my current project and has shown me the effect sizes, surface plots, and visualizations to help me understand where this suggested experiment is in the current space.
