Étude de casRecommencé
À propos du projet
Recommenu est une application de découverte culinaire alimentée par l'IA, conçue pour aller au-delà des recherches de restaurants génériques en recommandant à la fois où manger et quoi commander. L'application crée un profil gustatif personnalisé pour chaque individu, en apprenant ses préférences de saveurs et en prédisant dans quelle mesure il pourrait aimer un plat particulier. Les recommandations ne se limitent pas aux individus : Recommenu répond également au défi courant des repas en groupe en analysant plusieurs profils gustatifs (par exemple, des amis, des collègues ou des équipes) et en suggérant des restaurants proposant des plats qui maximisent la satisfaction collective. L'objectif était de créer une solution fluide et optimisée par les données qui personnalise les expériences culinaires tant au niveau individuel qu'au niveau du groupe.
Client Requirements
The project required a solution that could capture and analyze user taste data from diverse sources, build a robust recommendation engine capable of predicting restaurant and menu preferences, and support group dining decisions by merging multiple profiles into optimized results. Additionally, it was essential to automate restaurant and menu data collection at scale, leverage advanced AI and machine learning technologies for accuracy, and deliver everything within a scalable, user-friendly application framework.
Solutions We Delivered
To address these needs, we built a complete recommendation ecosystem powered by modern AI and machine learning. We began by implementing web scraping with ScraperAI and ScraperAPI to gather restaurant and menu data at scale. The raw data was then structured and enriched using ChatGPT, which also served as an alternate recommendation logic layer when additional insights were needed. At the core of the platform, we developed a sophisticated Recommendation Engine in Python, combining multiple technologies to ensure precision and scalability. Pinecone was integrated as a vector database to handle similarity searches and user profile matching, while Google Vertex AI provided the infrastructure for training and deploying machine learning models. For advanced personalization and prediction, we utilized Google Gemini AI, enabling the system to adapt dynamically to evolving user preferences. This integrated approach allowed us to create a personalized, intelligent, and scalable recommendation system that not only offers highly relevant restaurant and menu suggestions but also optimizes dining choices for groups. The result is a platform that transforms everyday dining decisions into data-driven experiences tailored to individual and collective tastes.
Client
Project Links
Country
US
Technologies we used
AWS
FastAPI
LangChain
LanGraph
Pinecone
Python
React Native
Redis
