Kaikaku's Epicure models split flavor pairing by chemistry vs recipes
kaikaku.ai released epicure, three models that separate ingredient pairings by recipe context vs flavor chemistry to allow precise culinary control.
startup kaikaku.ai introduced 'epicure,' a trio of models that isolate whether ingredient pairings are derived from shared flavor molecules or recipe frequency. testing showed the chemistry-based 'chem' model uniquely identified flavor relatives like chicken and pork or basil and oregano, whereas the recipe-based 'cooc' model suggested culinary staples like garlic and parmesan. the system processes 4.14 million recipes in seven languages, cleaned via llm embeddings, to map 1,790 ingredients. this allows for directional searches, such as shifting rice toward south asian flavors to discover curry leaf or fenugreek.