Sunthetics ML
Sunthetics ML is ann easy-to-use machine-learning (ML) platform capable of accelerating chemical process optimization with small datasets. Leveraging a combination of chemical engineering concepts and ML predictive algorithms, we can use very few data points to quickly predict optimal formulations for enhanced performance in new product applications.
Our software can improve existing manufacturing processes identifying opportunities for unprecedented efficiencies and it can be further used to identify anomalies and facilitate diagnostics.
- Shorter experimental campaigns - saves money, time, energy, and materials from lower number of experiments!
- Reaction agnostic approach
- Easy visualization and exploration of complex reaction trends
-Unlocks unprecedented performance
-No expertise in ML or statistics required
-No minimum number of experiments required
Sunthetics ML tool has been tested with chemical companies and academic research groups, showing that users can find optimal reaction conditions for enhanced reaction performance and material properties using up to 5 times less experiments than more traditional experimental campaigns
FAQ
- Does the tool work only with electrochemical reactions?
No! our platform learns from the reaction at hand. Has been tested in the optimization of electrochemical and non-electrochemical reactions, as well as the prediction of material properties!
- What type of variables does it take?
Our current platform takes 2-5 numerical variables (reaction parameters) and maximize one target output variable. New features for higher dimensionality and the use of categorical features are under development!
- Do I need to know AI or statistics to use this?
No! Our tool is extremely easy to use. You only upload an excel or csv file with the data from a few experiments and our algorithms will do all the work for you!
- Do I need hundreds of data points to use ML?
No! Our tool is designed to leverage information from very small datasets. Users commonly start using our tool after running only 4-5 experiments (4-5 data points)
