Showcase image

Model Building Toolkit

Utilize our universal molecular encoder and build custom predictive models with deep chemical understanding. Follow this 5-step process to transform your lab data into AI-powered formulation optimization.

Select Industry Example:

Our platform supports formulation development across diverse industries including pharmaceuticals, agriculture, food & beverage, personal care, and more.

1

Choose Data Type

Select the type of model that matches your prediction goal. Different data types enable different kinds of predictions. Anything with data that fits in these categories can be predicted.

Classification

Predict from 2 or more classes: true/false, categories, or scores/ranks

Examples:

Stability Adhesion

Regression

Predict a numerical value with continuous output

Examples:

Film Thickness pH Value

Time Series

Predict numerical values on a graph over time or conditions

Examples:

Viscosity Surface Tension
2

Set Target Criteria

Define what property you're predicting and the classes or range of values you expect. For this example, we chose Classification.

Example: Cross-Hatch Adhesion (ASTM)

Classification model with 6 adhesion grade classes

0B 1B 2B 3B 4B 5B
3

Upload Your Data

Model training starts automatically once you upload your formulation data with ingredients, conditions, and target property measurements.

Sample pH Acrylates Copolymer Titanium Dioxide HASE Temp (C) Adhesion
Sample 1 7.5 20% 8% 1% 25 0B
Sample 2 7.5 22% 9% 1% 25 1B
Sample 3 7.5 25% 10% 1.2% 25 2B

Data shown for illustrative purposes. Training starts automatically after upload.

4

Select Optimization Criteria

Choose ingredients for your new formulation, set concentration ranges, and define optimization targets for multiple properties.

Selected Ingredients

Acrylates Copolymer
Titanium Dioxide
HASE
Sodium Polyacrylate

Concentration ranges and values for each ingredient can be specified to constrain the optimization space.

Optimization Targets

Viscosity Target: 3 Pa·s
Surface Tension Target: 35 mN/m
Cross-Hatch Adhesion Target: 4B or 5B
Film Thickness Target: 50 µm
Stability Target: True
Cost Target: Minimize
5

Get Ranked List of Best Potential Concentrations

The optimizer runs dozens of simulations to find the best possible compositions. Get a ranked list to do informed trials instead of trial-and-error.

91.4
Best Score
150
Configurations
#1
Composite Score: 91.4
Viscosity: 3.04 Pa·s | Surface Tension: 34 mN/m | Stability: True | Adhesion: 4B | Film Thickness: 50.00 µm | Cost Ranking: 1
Acrylates Copolymer
29.360 %
HASE
1.435 %
Titanium Dioxide
8.434 %
Sodium Polyacrylate
0.285 %
#2
Composite Score: 89.3
Viscosity: 2.89 Pa·s | Surface Tension: 30 mN/m | Stability: True | Adhesion: 4B | Film Thickness: 48.78 µm | Cost Ranking: 3
Acrylates Copolymer
21.864 %
HASE
1.598 %
Titanium Dioxide
10.972 %
Sodium Polyacrylate
0.463 %
#3
Composite Score: 87.7
Viscosity: 2.89 Pa·s | Surface Tension: 33 mN/m | Stability: True | Adhesion: 4B | Film Thickness: 50.54 µm | Cost Ranking: 2
Acrylates Copolymer
21.561 %
HASE
1.256 %
Titanium Dioxide
7.780 %
Sodium Polyacrylate
0.245 %

Use these ranked formulations to guide your lab experiments with higher confidence

1

Choose Data Type

Select the type of model that matches your prediction goal. Different data types enable different kinds of predictions. Anything with data that fits in these categories can be predicted.

Classification

Predict from 2 or more classes: true/false, categories, or scores/ranks

Examples:

Stability Compatibility

Regression

Predict a numerical value with continuous output

Examples:

Zein Value pH Value

Time Series

Predict numerical values on a graph over time or conditions

Examples:

Viscosity Foam Longevity
2

Set Target Criteria

Define what property you're predicting and the classes or range of values you expect. For this example, we chose Classification.

Example: Texture

Classification model with 5 texture classes

Smooth Watery Creamy Heavy Gel-Like
3

Upload Your Data

Model training starts automatically once you upload your formulation data with ingredients, conditions, and target property measurements.

Sample pH SLES Cocamidopropyl Betaine Glycerin PEG-7 Glyceryl Cocoate Fragrance Temp (C) Texture
Sample 1 5.5 14% 8% 3% 1.50% 0.30% 25 Smooth
Sample 2 5.5 10% 3% 1% 0.50% 0.20% 25 Watery
Sample 3 5 16% 4% 3% 2.50% 0.50% 25 Creamy

Data shown for illustrative purposes. Training starts automatically after upload.

4

Select Optimization Criteria

Choose ingredients for your new formulation, set concentration ranges, and define optimization targets for multiple properties.

Selected Ingredients

Sodium Lauryl Ether Sulfate
Cocamidopropyl Betaine
Glycerin
PEG-7 Glyceryl Cocoate
Fragrance

Concentration ranges and values for each ingredient can be specified to constrain the optimization space.

Optimization Targets

Viscosity Target: 5 Pa·s
Stability Target: True
Texture Target: Smooth
Zein Value (Irritability) Target: Minimize
Lubricity Target: Med-High
Cost Target: Minimize
5

Get Ranked List of Best Potential Concentrations

The optimizer runs dozens of simulations to find the best possible compositions. Get a ranked list to do informed trials instead of trial-and-error.

91.6
Best Score
150
Configurations
#1
Composite Score: 91.6
Viscosity: 5.06 Pa·s | Stability: True | Zein Value: 103.49 | Lubricity: Med-High | Texture: Smooth | Cost Ranking: 2
Sodium Lauryl Ether Sulfate
13.983 %
Cocamidopropyl Betaine
4.580 %
Glycerin
2.132 %
PEG-7 Glyceryl Cocoate
1.938 %
Fragrance
0.308 %
#2
Composite Score: 88.8
Viscosity: 5.16 Pa·s | Stability: True | Zein Value: 86.81 | Lubricity: Med-High | Texture: Smooth | Cost Ranking: 3
Sodium Lauryl Ether Sulfate
19.210 %
Cocamidopropyl Betaine
4.457 %
Glycerin
1.103 %
PEG-7 Glyceryl Cocoate
1.357 %
Fragrance
0.407 %
#3
Composite Score: 86.1
Viscosity: 5.23 Pa·s | Stability: True | Zein Value: 72.75 | Lubricity: Med-High | Texture: Smooth | Cost Ranking: 1
Sodium Lauryl Ether Sulfate
19.931 %
Cocamidopropyl Betaine
7.689 %
Glycerin
4.534 %
PEG-7 Glyceryl Cocoate
1.285 %
Fragrance
0.198 %

Use these ranked formulations to guide your lab experiments with higher confidence

1

Choose Data Type

Select the type of model that matches your prediction goal. Different data types enable different kinds of predictions. Anything with data that fits in these categories can be predicted.

Classification

Predict from 2 or more classes: true/false, categories, or scores/ranks

Examples:

Shelf Stability Texture

Regression

Predict a numerical value with continuous output

Examples:

Water Activity pH Value

Time Series

Predict numerical values on a graph over time or conditions

Examples:

Viscosity Color Stability
2

Set Target Criteria

Define what property you're predicting and the classes or range of values you expect. For this example, we chose Classification.

Example: Shelf Stability

Classification model with 2 stability classes

Unstable Stable
3

Upload Your Data

Model training starts automatically once you upload your formulation data with ingredients, conditions, and target property measurements.

Sample pH Soybean Oil Distilled White Vinegar Sugar Ground Mustard NaCl Lecithin Temp (C) Shelf Stability
Sample 1 3.6 80% 4% 1% 1% 1% 4% 25 Unstable
Sample 2 3.6 80% 4% 1% 1% 1% 7% 25 Stable
Sample 3 3.6 80% 4% 1% 1% 1% 10% 25 Stable

Data shown for illustrative purposes. Training starts automatically after upload.

4

Select Optimization Criteria

Choose ingredients for your new formulation, set concentration ranges, and define optimization targets for multiple properties.

Selected Ingredients

Soybean Oil
Distilled White Vinegar
Sugar
Ground Mustard
NaCl
Lecithin

Concentration ranges and values for each ingredient can be specified to constrain the optimization space.

Optimization Targets

Shelf Stability Target: Stable
Texture Target: Creamy
Water Activity Target: 0.92
pH Value Target: 3.6
Viscosity Target: 3.5 Pa·s
Color Stability Target: Stable Over 6 Months
5

Get Ranked List of Best Potential Concentrations

The optimizer runs dozens of simulations to find the best possible compositions. Get a ranked list to do informed trials instead of trial-and-error.

94.7
Best Score
150
Configurations
#1
Composite Score: 94.7
Shelf Stability: Stable | Texture: Creamy | Water Activity: 0.918 | pH: 3.62 | Viscosity: 3.485 Pa·s | Color Stability: Stable
Soybean Oil
79.847 %
Distilled White Vinegar
4.123 %
Sugar
0.985 %
Ground Mustard
1.034 %
NaCl
0.976 %
Lecithin
8.265 %
#2
Composite Score: 92.3
Shelf Stability: Stable | Texture: Creamy | Water Activity: 0.924 | pH: 3.58 | Viscosity: 3.612 Pa·s | Color Stability: Stable
Soybean Oil
79.624 %
Distilled White Vinegar
4.056 %
Sugar
1.012 %
Ground Mustard
0.963 %
NaCl
1.028 %
Lecithin
9.487 %
#3
Composite Score: 90.1
Shelf Stability: Stable | Texture: Creamy | Water Activity: 0.911 | pH: 3.64 | Viscosity: 3.293 Pa·s | Color Stability: Stable
Soybean Oil
80.213 %
Distilled White Vinegar
3.967 %
Sugar
1.021 %
Ground Mustard
0.989 %
NaCl
1.045 %
Lecithin
6.783 %

Use these ranked formulations to guide your lab experiments with higher confidence

Ready to transform your formulation process?

See how FastFormulator can help you reduce development time, minimize failed experiments, and bring products to market faster with AI-powered predictions.