Turbo Till

Turbotill is a real-time AI order-taking agent powered by Gemini Live API that helps restaurants serve customers faster without depending on staff availability.

Comment

Inspiration

Restaurants and food businesses often struggle with one simple but critical task: taking customer orders quickly and consistently.

Small restaurants usually cannot afford multiple waiters or staff members to manage terminals, tables, and phone orders at the same time. During busy hours, even larger businesses experience bottlenecks when staff are unavailable or temporarily away from the counter.

We realized that most AI systems today are limited to text chatbots, but real restaurant operations require live, natural conversations with customers.

This inspired us to build Turbo Till, a real-time AI order-taking system powered by Gemini Live, where customers can simply talk to an AI agent and place orders naturally — just like speaking to a human staff member.

Our goal was to explore how live AI agents can assist real-world business operations, not just conversations.

What it does

Turbo Till is a real-time AI order-taking platform that allows restaurants to deploy live conversational AI agents to interact with customers and place orders automatically.

Using the Gemini Live Agent API, the AI agent can:

  • Talk with customers in real time
  • Understand food orders naturally
  • Handle interruptions during conversation
  • Confirm order details
  • Send order data to external systems

Businesses can create different types of AI agents, including:

  • Terminal Agents – for front-desk or kiosk ordering
  • Table Order Agents – for dine-in ordering
  • Drive-Thru Agents – for vehicle-based ordering
  • Embedded POS Agents – integrated inside existing web POS systems

A key feature of Turbo Till is the Human + AI hybrid workflow.

If a human staff member needs to step away from the counter, they can activate the AI agent, and the AI immediately continues serving the customer without interrupting the ordering process.

Once the order is confirmed, Turbo Till sends the order data to external systems using webhooks, ensuring that the restaurant’s POS or backend systems remain synchronized.

How we built it

Turbo Till was built as a real-time AI agent system using Google Gemini Live and Google Cloud infrastructure.

Frontend

  • React interface for interacting with the AI agent in real time

AI Agent

  • Gemini Live API for real-time conversational interaction

Backend

  • Node.js server for agent orchestration and order processing
  • Webhook system for sending confirmed orders to external POS systems

Cloud Infrastructure

  • Hosted on Google Cloud
  • Uses Gemini models for conversational understanding and reasoning

The system processes customer speech in real time, extracts structured order information, and sends it to external systems once the order is confirmed.

Challenges we ran into

Building a real-time conversational ordering system came with several challenges.

One major challenge was designing natural order-taking conversations that feel smooth and realistic. Customers often interrupt themselves, modify their order mid-conversation, or add additional items.

Another challenge was converting unstructured voice input into structured order data that can be reliably processed by restaurant systems.

We also needed to ensure low-latency real-time responses so the interaction felt natural for customers.

Finally, integrating the AI ordering workflow with external POS systems required designing a flexible webhook-based architecture so Turbo Till could work with different restaurant software.

Accomplishments that we're proud of

We are proud that Turbo Till demonstrates how AI agents can become operational assistants in real-world business environments.

Key accomplishments include:

  • Successfully building a real-time AI order-taking experience using Gemini Live
  • Creating a human + AI hybrid workflow for restaurant operations
  • Designing a system that integrates with external POS systems through webhooks
  • Demonstrating how conversational AI can move beyond chatbots and support real business workflows

The project highlights the potential of AI agents to help businesses remain operational even when human staff are temporarily unavailable.

What we learned

While building Turbo Till, we gained valuable insights into designing real-time AI agents for operational environments.

We learned how to:

  • Design interruptible natural conversations using Gemini Live
  • Convert spoken customer requests into structured order data
  • Build AI workflows that collaborate with human staff
  • Integrate conversational AI with existing business systems
  • Deploy scalable AI services using Google Cloud

One of the biggest takeaways was understanding how live AI agents can move beyond simple chatbots and become real operational tools for businesses.

What's next for Turbo Till

We plan to expand Turbo Till into a more advanced AI-powered restaurant automation platform that helps businesses manage the entire ordering experience with intelligent AI agents.

Future improvements include:

  • Multilingual ordering support so customers can place orders in their native language
  • Menu image recognition using Gemini multimodal capabilities, allowing customers to scan a menu with their camera and place orders instantly
  • AI-powered upselling and recommendations where the agent intelligently suggests combos, add-ons, or popular items
  • Analytics dashboards providing restaurants with insights into ordering patterns, peak hours, and customer preferences
  • Native integrations with major POS platforms to simplify deployment for restaurants
  • Drive-thru camera support where the AI can interact with customers through audio and visual context
  • WhatsApp Call Agent that can receive customer calls on WhatsApp, talk with them using voice, and automatically place orders
  • Vision-based ordering assistant where customers can show the AI an item or menu board and the system understands what they want
  • AI kitchen intelligence that predicts demand and helps restaurants prepare popular items ahead of time

Our long-term vision is to build a platform where AI agents work alongside human staff to automate ordering, improve service speed, and create smarter restaurant operations worldwide.

Built With

Share this project:

Updates

Submission history