Inspiration

Global businesses struggle to understand customer feedback across different languages and cultures. A 5-star review in Japan might actually indicate dissatisfaction due to cultural politeness, while direct criticism from German customers could be constructive feedback. We built Jback to bridge this cultural intelligence gap using real-time streaming and AI.

What it does

Jback is a Cross-Cultural Feedback Intelligence Platform that:

  • Collects customer feedback in 100+ languages via QR codes and web forms
  • Streams feedback in real-time through Confluent Cloud Kafka
  • Analyzes sentiment with cultural context awareness using Google Gemini AI
  • Provides an AI chat agent that understands cultural nuances and delivers actionable business insights
  • Translates and preserves original meaning while explaining cultural communication styles

How we built it

Layer Technology
Frontend Next.js 14, React, TailwindCSS, shadcn/ui
AI Engine Google Gemini 2.0 Flash via Vercel AI SDK
Real-time Streaming Confluent Cloud Kafka
Database TiDB Serverless + Prisma ORM
Authentication NextAuth.js
Deployment Vercel Edge Functions

Challenges we ran into

  • Handling cultural context in sentiment analysis — same words mean different things across cultures
  • Migrating from legacy AI SDK to the new @ai-sdk/google package with breaking API changes
  • Implementing proper streaming responses compatible with React hooks
  • Managing real-time Kafka consumers in serverless environment
  • Balancing translation accuracy while preserving cultural nuances

Accomplishments that we're proud of

  • Built a working AI chat agent that actually understands cultural differences in feedback
  • Achieved real-time feedback streaming with sub-second latency using Confluent Cloud
  • Created an intuitive QR-based feedback collection system that works globally
  • Successfully integrated Google Gemini for both analysis and conversational AI
  • Developed cultural intelligence features that explain why customers communicate differently

What we learned

  • Cultural context is crucial for accurate sentiment analysis — direct translation isn't enough
  • Confluent Cloud's managed Kafka significantly simplifies real-time streaming architecture
  • Google Gemini excels at multilingual understanding and cultural nuance detection
  • Streaming AI responses dramatically improves user experience in chat interfaces
  • Building for global audiences requires thinking beyond just language translation

What's next for Jback

  • Voice Feedback — Add speech-to-text for verbal feedback in native languages
  • Predictive Analytics — Use historical patterns to predict customer churn by region
  • Custom Cultural Profiles — Let businesses define their target market's cultural expectations
  • Integration Hub — Connect with CRM systems, Slack, and support ticketing platforms
  • Advanced Reporting — Generate automated weekly cultural insight reports
  • Mobile App — Native iOS/Android apps for on-the-go feedback management

Built With

  • apache-kafka
  • confluent-cloud
  • google-gemini-ai
  • next.js
  • nextauth.js
  • prisma-orm
  • react
  • tailwindcss
  • tidb-serverless
  • typescript
  • vercel
  • vercel-ai-sdk
Share this project:

Updates