Our Technology

Why machine learning?

‍ Optimizing chemical processes, reactions, or formulations is often a trial-and-error process guided by expert intuition. But when you’re dealing with many variables and complex goals, it’s easy to head in the wrong direction—wasting valuable time and resources.

Machine learning accelerates this process by revealing complex relationships between variables, predicting outcomes, and guiding you toward the most promising conditions—faster and with fewer experiments.

A platform technology

Our platform is reaction agnostic, meaning we can enter a variety of verticals in the chemical industry and have a large impact in a short amount of time. Whether it's pharmaceuticals, specialty chemicals, or anything in between, we can help accelerate innovation.

The platform is accessible via internet browser and doesn't require any machine learning knowledge from the user, making it truly easy-to-use.

SuntheticsML is used across sectors and scales. Typical applications include:

Reaction understanding and engineering (e.g., modeling of reaction behavior, parameter effects and interactions)

Process characterization (e.g., prediction of failure zones, identification of safe operation ranges for manufacturing)

Process development and optimization

Formulation optimization

Analytical method development and optimization

Process scale-up (parameter tuning and optimization)

SuntheticsML has been successfully implemented across a broad range of scientific fields and industries.

Applications include catalysis, electrochemistry, photochemistry, mechanochemistry, biocatalytic cascades, separation processes, analytical chemistry, enzymatic reactions, biologics, and both organic and inorganic chemistry. These implementations span industries such as pharmaceuticals, specialty and basic chemicals, food and beverages, advanced materials, cosmetics, personal care, and more.

Technology features

Process characterization

Unlocks unprecedented performance

No expertise in ML or statistics required

Only 5 data points required to begin analysis

Easy visualization and exploration of complex reaction trends

In small molecule development, crystallizations, biocatalysts, electrochemistry, catalytic processes, formulations, biologics, and more.

Case studies

Electrochemical Reaction Optimization

Performance Improvement:

Sunthetics identified a 7% higher faradaic efficiency, optimizing flow rate, current density, and reactant concentration.

Data reduction:

Sunthetics' campaign used 75% less experiments, time, and resources.

Total number of experiments required: 9

Controlling aspect ratio of crystals

Performance Improvement:

Sunthetics identified a 13% better aspect ratio optimizing 5 crystallization variables.

Data reduction:

Sunthetics' campaign used 70% less experiments, time, and resources than the company's Design of Experiments.

Total number of experiments required: 5

Electrochemical Reaction Optimization

Performance Improvement:

Sunthetics identified a 40% higher fracture toughness in polymer double networks optimizing formulation and polymerization conditions.

Data reduction:

Sunthetics' campaign used 80% less experiments, time, and resources.

Total number of experiments required: 4

FAQ

Is this the same as statistical design of experiments (DoE)?

SuntheticsML is based on a bayesian optimization approach that guides experimentation to reduce the number of experiments required. Statistical DoE can have limitations when modeling complex reactions that have multiple local minima/maxima. SuntheticsML can map complex reaction spaces more effectively than traditional DoE, achieving higher performance with fewer experiments. SuntheticsML is a stand-alone tool capable of improving scientists' productivity, guiding experimentation, and achieving unprecedented efficiencies. However, users can always complement their DoE-enabled analysis with our tool to enhance the accuracy of the models and predictions.

What if I do not wish to use the cloud-based software?

Our cloud-based software is designed to be an AI partner available to the researcher at all times. However, if you do not wish to use the software yourself, we can assign one of our engineers to your project. They will run all optimizations for you and assist you through the process.

Do I need to know AI, programming, or statistics to use this?

No, our tool is extremely easy to use! Users upload an Excel or CSV file with the data from a few experiments and our algorithms will do all the work.

Is my data safe?

Absolutely! We follow industry best practices to ensure your data is protected. You can visit our Trust Center to see the latest compliance updates. We conduct annual penetration tests and are currently SOC 2 Type II certified.

Do I need hundreds of data points to use ML?

No, SuntheticsML is designed to leverage information from very small datasets, which differentiates us from other optimization platforms. Users commonly start using our algorithms with only 5 data points.

What types of reactions does the tool work with?

Any kind! SuntheticsML is built to be reaction-agnostic. Our tool learns from the reaction/system at hand—the initial dataset you provide and subsequent experiments that the algorithms suggest. Our algorithms have been used with electrochemical reactions, catalytic processes, formulation optimization, prediction of material properties, mechanochemistry, photochemistry, and much more. Look at our summary of case studies and contact us if you have questions about a particular research area!

Why are the algorithms so accurate?

ML predictive algorithms are as accurate as the quality of the data provided for model training. However, Sunthetics’ ML algorithms autocorrect themselves. We learn from your initial data and suggest subsequent experiments designed to confirm, adapt, and correct predicted trends. The data you collect is smartly selected to fast-track the mapping of outliers and hidden optimal areas of operation.

What are Sunthetics’ machine-learning (ML) algorithms?

ML is an approach designed to identify patterns in data and make predictions of behavior under untested conditions. At Sunthetics, we leverage proprietary algorithms in a modified Bayesian Optimization approach to learn from your data in the most efficient way. Our algorithms run millions of scenarios representing potential experiments to guide you through the optimization path. In other words, it saves you time and resources, minimizing frustration in R&D! We take your data, suggest 1+ subsequent experiment(s) to map your system, and achieve your optimization goals up to 32x faster!