Epicure
    KitchenCollectionMethodMCP
    Enter kitchen
    The intelligence behind Epicure

    Taste has a topology.

    Epicure maps relationships between ingredients, recipes and cultures into a navigable model of culinary possibility.

    Enter the flavour atlasRead the system
    Epicure ingredient relationship graph

    Pattern, not prescription.

    The map suggests direction. The cook keeps authorship.

    Three connected systems

    From research graph to a recipe you can cook tonight.

    Flavour Graph

    A research lineage that turns millions of culinary relationships into navigable space.

    Ingredient atlas

    A real 3D UMAP projection of Epicure's 1,790 ingredient embeddings. Drag to rotate, pinch or scroll to zoom.

    Loading the ingredient atlas

    Preparing projection

    The origins · 2011

    Research into the cultural diversity of cooking asked a deceptively simple question: are there universal patterns in ingredient combinations?¹

    FlavorGraph · 2021

    A deep-learning model trained on more than one million recipes and chemical-compound data revealed how ingredients interact in practice.² ³

    Epicure · now

    We make that latent map tangible, surfacing familiar harmonies and surprising pairings that still make culinary sense.

    Macrodata refinement

    Raw relationships become precise, contextual culinary intelligence.

    Signal refinement

    Noise to context
    PAIRINGCULTUREDIETRAW DATAREFINECULINARY SIGNALS

    Pattern recognition

    Semantic patterns are consolidated into a coherent flavour-pairing dictionary.

    Context enrichment

    Dietary, cultural and ingredient-category signals add the context a chef actually needs.

    Living intelligence

    Search, generation and recommendations share one continuously useful food model.

    Food intelligence

    The graph becomes an active co-chef, from first ingredient to finished plate.

    Flavour direction

    A composed dish vector
    BRIGHTAROMATICDEEPSWEETUMAMIDISHVECTOR

    Pairing direction

    Algorithms select high-value relationships while keeping the result diverse and cookable.

    Recipe composition

    AI develops culturally aware recipes across complexity, flavour and dietary constraints.

    Complete output

    Nutrition, visual direction and professional food imagery complete the cooking experience.

    Source material

    Read the research.

    The invention of flavour networks

    Flavor network and the principles of food pairing

    Current FlavorGraph research

    FlavorGraph: a large-scale food pairing model

    Open research model

    Explore FlavorGraph on GitHub

    Epicure by Kaikaku

    Cook beyond the obvious.

    KitchenRecipe collectionThe methodMCP for agents
    Kaikaku AI

    Protected by Cloudflare Turnstile. Cloudflare's Privacy Policy and Terms apply.

    Pinned AI models run through OpenRouter with provider data collection disabled. Read OpenRouter's Privacy Policy.

    PrivacySupportGitHubXLinkedInInstagram© 2026 Epicure