AI-Powered Recommendation & Matching Platform for F&B Engagement Growth

SmartDev partnered with a leading F&B company in Japan to boost engagement and revenue without increasing marketing spend. By deploying a secure, modular AI recommendation and matching engine across web and mobile platforms, the solution enabled personalized user matching, event recommendations, and AI-generated drink recipes. The result was a more engaging, scalable, and data-driven platform that improved conversion and long-term growth.
인원수

5 headcounts (1 Tech Lead, 2 Backend Developers, 1 Mobile Developer, 1 QA Engineer)

산업

Food & Beverage

제품 및 서비스

AI Recommendation & Matching Systems

시간 척도

Since 2025
(전진)

국가

일본

사업 개요

The client is a large-scale food and beverage enterprise in Japan with a strong presence across beverages and consumer food products. With an extensive customer base and a well-established brand, the company places high importance on customer engagement, brand experience, and continuous innovation in a highly competitive market.

As consumer interactions increasingly shift to digital channels, the client aims to enhance its web and mobile platforms through AI-driven personalization. The goal is to deepen user engagement, enable more relevant product and event discovery, and unlock new revenue opportunities, while maintaining enterprise-grade security, scalability, and operational reliability.

도전 과제

  • Authentication integration: Synchronizing Firebase Auth with the Supabase database proved difficult due to differing access control mechanisms.
  • UI/UX bugs: The app experienced complex frontend issues such as incorrect session ordering, inability to view user profiles, and form data being lost during screen transitions.
  • Notifications & email: Choosing a reliable service and integrating stable email and push notification flows was challenging (the requirement was to ensure important messages are never missed).
  • Data security & governance: There was a need to handle user data securely, comply with security standards, and be audit-ready for the AI system.

솔루션

  • Modular architecture: Built a separate recommendation/matching engine that is plug-and-play, allowing staged rollouts and clear measurement of results.
  • Authentication management: Integrated Firebase Auth with Supabase for shared user storage and access control. All API requests to AI services are authenticated via JWT tokens, combined with Supabase RLS to ensure each user can only access their permitted data.
  • UI/UX improvements: Upgraded web and mobile interfaces. Components now support real-time predictions, chat with a virtual assistant, and analytics dashboards. Fixed issues related to screen transitions and form persistence. These improvements make the user experience more intuitive and consistent.
  • Notification system: Combined SendGrid for system emails (verification, promotions, etc.) with FCM for in-app push notifications. Optimized configurations to reduce delivery failure rates and ensure users receive timely messages.
  • Enhanced security: Conducted a comprehensive review of Supabase database structure and implemented RLS policies. Prepared security guidelines and audit artifacts for third-party inspection. This ensures compliance with personal data protection standards and builds client trust.

당사의 기술 스택

반응하다
Material-UI
타입스크립트

Core functions

Matching Engine

The system uses AI to analyze user preferences and behavior from interaction history and user profiles. A OpenAI text-embedding-large model converts information into vectors. The engine matches users with similar tastes based on vector proximity.

Event matching

Similarly, the system computes and suggests events (meetups, mixology classes, etc.) based on user location and interests. For example, if a user likes fruity cocktails and is in Tokyo, the system will prioritize related events.

Personalized drink recipe generation

An AI component automatically generates drink recipes based on available ingredients and user tastes. This feature encourages repeat app usage as users discover new drinks tailored to them.

GPT-4o and similarity analysis

GPT-4o is used to analyze and explain why the system matched two users or matched a user to an event. This transparency increases user trust in the recommendations.

Advanced recommendation algorithms

Unlabeled data (e.g., scattered interaction logs) is processed with feature-extraction techniques and combined with LLMs to build an accurate recommendation dataset.

Outcomes

Increased revenue & engagement

Data indicates recommendation systems can boost conversion rates by up to 150%, and 91% of consumers are more likely to buy when they receive personalized suggestions.

Improved user experience

Intuitive UX and intelligent features increase customer satisfaction. 74% of consumers say AI improves their shopping experience.

Stronger security

Applying RLS and authenticated tokens protects personal data. This meets compliance expectations and builds trust.

Reliable notifications

SendGrid and FCM improve communication reliability, fewer email failures and notification misses. Users receive promotions and reminders in real time, increasing app usage frequency.

Control & measurement

Modular deployment lets SMBs monitor KPIs: engagement rate, conversion. AI can continuously optimize algorithms based on collected data to keep improving results.

시작하기

생각하고 있는 프로젝트가 있나요?

With SmartDev, you gain access to a wide range of expertise to seamlessly migrate your systems and optimize your cloud infrastructure. Share your vision with us, and we’ll respond by the next business day. Our team is ready to help your business unlock new opportunities and drive innovation.

오늘 시작해 볼까요?

뉴스레터를 구독하세요

SmartDev와 함께 기술 업계를 선도하세요! 최신 IT 뉴스, 업데이트, 인사이트를 받아보시려면 뉴스레터를 구독하세요.