Zara

• 

  Screenshot

Inspiration

Customer support today is slow, repetitive, and dependent on humans. Even simple tasks like refunds or cancellations take time.

We asked: Why can’t support systems take action instead of just replying?
That’s what led to Zara.

What it does

Zara is an AI voice agent that resolves customer issues in real time.

• Answers calls
  
• Understands intent
  
• Executes actions (cancel, refund, track)
  
• Confirms instantly
  

No tickets. No waiting. Just resolution.

How we built it

• Backend: Node.js + Express
  
• Database: SQLite
  
• AI: Gemini 1.5 Flash
  
• Voice: Twilio
  
• Frontend: React dashboard
  

Zara uses structured outputs (intent → action → response) to reliably execute tasks.

Challenges we ran into

• Ensuring safe execution of real actions
  
• Maintaining low latency in voice calls
  
• Handling multi-step conversations
  
• Making AI outputs consistent and reliable
  

Accomplishments

• Built a system that takes action, not just chats
  
• Real-time voice-based issue resolution
  
• End-to-end automated support workflow
  
• Working prototype with live dashboard
  

What we learned

• Users want outcomes, not conversations
  
• Structured AI = reliable AI
  
• Speed matters a lot in voice systems
  
• Guardrails are critical for real-world actions
  

What's next for Zara

• Multi-channel support (chat, WhatsApp)
  
• Deeper integrations (payments, CRM)
  
• Better accuracy and personalization
  
• Move toward fully autonomous support systems
  

Zara resolves. Not responds.

Built With

• elevenlabs
• express.js
• gemini
• react
• resend
• sqlite
• twilio