How It Works
Flavour Graph
Deep learning power
The Origins (2011)
The cultural diversity of culinary practice raised a fundamental question: are there universal patterns in ingredient combinations?1
FlavorGraph (2021)
After 10 years of research, FlavorGraph - a deep learning model trained on over 1M recipes and chemical compound data that reveals how ingredients interact and are paired in recipes - was published.2,3
Epicure (2025)
Using the FlavorGraph dataset, we can identify the usual, but also unusual flavour pairings and use them to create recipes that are more likely to be enjoyed.
Macrodata Refinement
AI-powered data processing
Pattern Recognition
AI identifies semantic patterns in the data and consolidates it into a comprehensive flavour pairing dictionary.
Data Processing
AI processes and analyzes the semantically consolidated data and enriches it with dietary and categorization information.
Epicure Magic
Epicure seamlessly exploits this data to create a comprehensive AI-assisted food generation tool.
Food Intelligence
AI-powered food generation
Smart ingredient selection
Smart algorithms select best flavour pairings from the FlavorGraph, and provide AI-assisted pairing recommendations.
Advanced recipe generation
The ingredients are then processed by AI to generate various types of recipes at different complexity levels with cultural awareness and professional-photography style image generation.
Nutritional information
The recipes are enriched with nutritional information and dietary information with AI-powered analysis.