Data-Driven Solutions · ML & Deep Learning
Algorithms that deliver results
Real-world use cases from food technology, medicine, and ecology — each model is individually tailored to your data.
All visualizations are based on synthesized demonstration data.
Request a Project
Gradient Boosting (XGBoost)
Allergen Risk Assessment in Product Formulations
Food Technology & Consumer Protection. XGBoost evaluates the cross-contamination risk in shared production lines. SHAP values explain each prediction individually — auditable for quality assurance and regulatory authorities.
ⓘ Adjust the production parameters — risk score and SHAP waterfall update automatically.
Level 3 — Standard CIP
40 % of runtime
SHAP Waterfall: Feature Contribution
Risk Score
0.58
Score 0–1
Classification
Medium
Confidence
81 %
Baseline Method Comparison: F1 Score
Decision Tree
Clinical Malnutrition Screening (MedTech)
Medical Technology & Clinical Decision Support. The tree replicates the validated NRS-2002 score. Since every branch is visible, the model meets the transparency requirements of the EU AI Act for high-risk medical devices.
22.0
4.0 %
Decision Path
Feature Importance (Gini)
Naïve Bayes · Text Classification
Automated Classification of Consumer Complaints
Consumer Protection & Regulatory Affairs. Naïve Bayes classifies incoming complaint texts into priority categories. Latency under 2 ms enables real-time triaging for thousands of reports daily.
Category Probabilities
Diagnostic Keywords
Precision
91.4 %
Recall
88.7 %
Latency
<2 ms
per document
Random Forest · Ecology
Classifying Microplastic Sources in Water Samples
Environmental Analytics & Ecology. A Random Forest classifies microplastic particles based on spectroscopic and morphological features by source of origin — without prior knowledge of the source composition.
ⓘ Water type determines the source profile. Trees and particle size show their effect on OOB accuracy.
100 Trees
500 µm
Feature Importance (Radar)
Class Distribution in Water Body
OOB Accuracy
92.3 %
Trees
100
Source Classes
6
Support Vector Machine (SVM)
Olive Oil Authentication via NIR Spectroscopy
Food Control & Fraud Detection. An SVM separates genuine olive oil from adulterated samples using NIR spectra. Adulterations detectable from ~10 % foreign oil content. Click into the diagram to test a sample.
ⓘ Blend ratio and temperature simulate real testing conditions and shift the decision boundary.
15 % Foreign Oil
25 °C
Decision Boundary (PCA)
→ Click = test your own sample
Precision
96.1 %
Recall
94.3 %
AUC-ROC
0.982
ROC Curve
Convolutional Neural Network (CNN) · Deep Learning
Histological Tissue Quality in Meat Products
Food Technology & Quality Assurance. A fine-tuned EfficientNet-B0 classifies histological cross-section images into four quality levels. The Grad-CAM map reveals which image regions drive the decision.
Grad-CAM Activation Map
Accuracy
94.2 %
F1 Score
0.937
Inference
38 ms
Confusion Matrix
Class Probability
All visualizations are based on realistically synthesized demonstration data. For your project, we develop models individually tailored to your needs.
Request a Project