Inspiration

Car rental customers often face long queues at checkout, a frustrating, time-consuming experience. While in-app checkouts help speed up the process, they miss a critical revenue driver: human-like upselling. Real sales agents excel at empathy, timing, and personalized recommendations, which static digital flows fail to replicate. We were inspired to close this gap. Using realistic AI avatars, we combine the warmth, trust, and conversational nuance of a human agent with the convenience of a seamless, queue-free, and instant mobile checkout.

What it does

SixthGear processes a customer’s full rental booking end-to-end. It:

  • Validates the driver’s license
  • Guides the customer through a fast digital checkout
  • Holds natural, friendly conversations about the customer’s trip
  • Recommends personalized upgrades (car class, protections, addons)
  • Explains each suggestion with helpful reasoning, just like a human sales agent The experience feels like talking to a human sales rep… but instantly, anywhere.

How we built it

We combined Beyond Presence’s hyper-realistic AI avatars with:

  • OpenAI Speech-to-Text and Text-to-Speech
  • LiveKit for real-time audio/visual interaction
  • GPT-based state management to transition between sales stages
  • A custom decision-making engine for protection and addon recommendations
  • A FastAPI backend connecting directly to Sixt’s booking APIs

Challenges we ran into

  • Surviving on 4 hours of sleep in 2 nights
  • Integrating LiveKit into our custom frontend
  • Keeping avatar responses natural while managing strict agent states
  • Getting consistent reasoning from LLMs despite their nondeterminism
  • Ensuring high-quality, human-like upsell conversations, maintaining speed and accuracy

Accomplishments that we're proud of

Pulling through, making it work finally after many struggles.

What we learned

  • Read the API Documentation carefully
  • Don't be too adventurous in using new libraries at a hackathon

What's next for SixthGear

  • Video integration and image segmentation to learn more about the customer visually and tailor the conversation to what we see.
  • Noise-cancellation for more robust service in busy environments like the airport.

Built With

Share this project:

Updates