KITI AI

Parses unstructured recipe text (from URLs, images, or plain text) and extracts a normalized ingredient list with quantities, units, and substitution mappings. Uses NLP-based entity recognition to identify ingredients, quantities, and preparation notes, then maps them to a canonical ingredient database for standardization across different recipe formats and culinary terminology variations.

Accepts user dietary profiles (allergies, restrictions, preferences, cuisines) and modifies extracted ingredient lists and recipes in real-time by substituting incompatible ingredients with alternatives, adjusting quantities, and filtering recipes that don't match constraints. Maintains a preference graph that learns from user selections and applies rules-based filtering with optional ML-based recommendation scoring.

Accepts a base recipe and target serving size, then scales all ingredient quantities proportionally while recalculating estimated costs based on real-time or cached pricing from meal kit partners. Uses dimensional analysis for unit conversion (cups to grams, etc.) and applies non-linear scaling rules for ingredients that don't scale linearly (spices, leavening agents, salt). Integrates with partner pricing APIs to show cost deltas for different serving sizes.

Converts personalized, scaled ingredient lists into delivery orders by matching ingredients to meal kit partner SKUs, handling inventory availability, and submitting orders through partner APIs or checkout flows. Manages order state (pending, confirmed, shipped) and coordinates with multiple meal kit providers (HelloFresh, EveryPlate, etc.) through standardized integration points, handling provider-specific ingredient substitutions and delivery constraints.

Integrates with recipe sources (food blogs, recipe databases, user uploads) and surfaces recipes that match user preferences, dietary restrictions, and available ingredients. May include web scraping, API integrations with recipe databases (Spoonacular, Edamam, etc.), or user-generated recipe uploads. Applies ranking/filtering based on user profile, cuisine preferences, and ingredient availability from meal kit partners.

Automatically enriches recipe data with structured metadata including cuisine type, dietary classifications (vegan, gluten-free, etc.), allergen information, cook time, difficulty level, and nutritional data. Uses NLP and rule-based extraction to infer metadata from recipe text, or integrates with third-party nutrition APIs (USDA FoodData Central, Nutritionix) to calculate nutritional profiles. Enables filtering and personalization downstream.

Tracks user interactions (recipes viewed, ordered, rated, skipped) and learns preference patterns to improve future recommendations and personalization. May use collaborative filtering (similar users' preferences), content-based filtering (recipe features), or hybrid approaches. Feedback loop allows users to rate recipes and adjust preferences, which updates recommendation models and personalization rules.

Aggregates ingredients from multiple recipes into a unified shopping list, deduplicates items, and optimizes for meal kit delivery by grouping ingredients by provider, delivery window, or cost efficiency. May suggest bulk purchasing or ingredient reuse across recipes to minimize waste and cost. Handles quantity aggregation (e.g., 2 cups flour from recipe A + 1 cup flour from recipe B = 3 cups total) and unit normalization.

Transforms traditional recipe instructions into meal-kit-friendly formats by breaking down steps into pre-portioned ingredient packs, adding timing synchronization for simultaneous cooking of multiple components, and simplifying instructions for users with varying cooking skill levels. May include video guidance generation or step-by-step photo instructions. Adapts instructions based on meal kit partner requirements (e.g., HelloFresh's specific format) and user cooking skill level.