The Food Intelligence Graph
The foundation layer powering restaurant AI. We are building the world's restaurant knowledge graph — connecting restaurants, menus, dishes, ingredients, cuisine, pricing, nutrition and customer preferences into one computable layer.
Seven dimensions of restaurant intelligence
Every entity in the food world — from a single ingredient to a global cuisine — is connected, typed and computable.
Restaurants
Venues, locations, hours, concepts and service models — structured and linked.
→ linked to menus, dishes & customersMenus
Full menu structures with sections, items, modifiers and pricing relationships.
→ semantically parsed, not just textDishes
Every dish classified by cuisine, cooking method, flavor profile and seasonality.
→ 100K+ dish types and growingIngredients
Ingredient ontology with substitutions, allergens, sourcing and seasonality.
→ 50K+ ingredient entitiesCuisine
A global cuisine taxonomy — regional styles, fusion traditions and culinary lineage.
→ 200+ cuisine categoriesPricing & Nutrition
Price intelligence, margin estimation and nutritional attributes per dish and ingredient.
→ real-time market benchmarksCustomer Preferences
What diners order, why they choose it, and how preferences shift across time, location and context — the demand layer that closes the loop.
→ powers personalization & forecastingWhat the graph makes possible
Understand any menu, instantly
Feed an unstructured menu into the graph. Get back structured dishes with cuisine tags, ingredient lists, allergen flags, price benchmarks and margin estimates — no manual data entry.
- Parse PDF, image or text menus in seconds
- Auto-classify dishes and detect cuisine
- Flag allergens and dietary attributes
{
"dish": "Margherita Pizza",
"cuisine": "Italian",
"ingredients": ["mozzarella", "tomato", "basil"],
"allergens": ["dairy", "gluten"],
"price_benchmark": "$12–$16",
"margin_estimate": "68%"
}
Power AI agents that know food
Your AI agent doesn't need to learn what "spicy" means or which wines pair with tomato sauce. The graph gives it restaurant-native reasoning from day one.
- Ingredient substitution and pairing logic
- Flavor profile reasoning and menu design
- Customer preference-aware recommendations
From unstructured menus to structured intelligence.
The Food Intelligence Graph is not hand-curated — it is continuously constructed by a four-stage pipeline that turns raw restaurant data into a living, computable knowledge base.
Ingest
Collect menus from POS systems, delivery platforms, restaurant websites and PDF/image uploads. Multi-format ingestion: text, structured JSON, OCR from images.
Parse & Normalize
NLP models extract dish names, descriptions, prices, modifiers and section structure. Normalize to canonical forms — "spaghetti bolognese" → "Spaghetti Bolognese" with cuisine=Italian.
Entity Link
Link each dish to the ingredient ontology, cuisine taxonomy and nutrition database. Resolve ambiguous entities ("pepper" → bell pepper vs. chili pepper) via context and confidence scoring.
Enrich & Learn
Compute derived attributes: margin estimates, price benchmarks, flavor profiles, dietary flags. Every new menu refines the models — the graph gets smarter with each ingestion cycle.
AVG PROCESSING TIME: 4.2 SECONDS PER MENU · 96.8% ENTITY LINKING ACCURACY
Intelligence you can trust.
A knowledge graph is only as good as its data quality. Every node and edge in the Food Intelligence Graph is scored, validated and governed — so your AI agents and applications reason on reliable foundations.
Confidence Scoring
Every entity link and attribute extraction carries a confidence score. Low-confidence nodes are flagged for human review or excluded from agent reasoning until validated.
- Per-entity confidence scores (0–1)
- Threshold-based filtering for agent use
- Human-in-the-loop review queue
- Confidence exposed via API metadata
Provenance & Lineage
Every graph node traces back to its source — which restaurant, which menu, which ingestion batch. When source data changes, dependent nodes are flagged for recomputation. Nothing in the graph is untraceable.
- Full data lineage per node and edge
- Source attribution and ingestion timestamps
- Incremental updates on source change
- Audit trail for all graph modifications
Continuous Validation
Automated quality checks run on every ingestion batch: schema validation, duplicate detection, cross-source consistency, outlier flagging. Data that fails validation is quarantined — never silently propagated into the graph.
- Automated schema and format validation
- Duplicate and conflict detection
- Cross-source consistency checks
- Quarantine pipeline for bad data
Ontology Governance
The food ontology — cuisines, ingredients, dish types, flavor profiles — is governed by a combination of expert curation and statistical learning. New entity types are proposed by models, reviewed by domain experts, and versioned for backward compatibility.
- Expert-curated core ontology
- Model-proposed entity extensions
- Versioned ontology with deprecation policy
- Domain expert review board
The world's cuisines, in one graph.
From street food to fine dining, from Tokyo to Toronto — the Food Intelligence Graph spans global cuisines, regional variations and culinary traditions. Coverage grows with every restaurant that joins.
From Italian and Japanese to regional Chinese, Mexican, Thai, Indian, Middle Eastern, African and fusion — every major cuisine and its sub-styles.
Produce, proteins, dairy, grains, spices, oils, sauces and specialty ingredients — with substitutions, allergens, seasonality and sourcing attributes.
Canonical dish types with cooking method, flavor profile, cuisine lineage, typical ingredients and common variations across regions.
Venues across 40+ countries, linked to their menus, dishes, service models and customer preference signals — growing daily.
North America, Europe, APAC, Latin America and Middle East — with localized cuisine taxonomies and regional ingredient ontologies.
Menu parsing in English, Spanish, French, German, Italian, Japanese, Chinese (Simplified & Traditional), Korean, Portuguese, Arabic and more.
Query the world's restaurant knowledge graph.
The Food Intelligence Graph is available via the Food Graph API and Menu Intelligence API. Start building restaurant-native AI today.