• 

  Interactive map

• 

  AI Video Chat

AI Real Estate Agent

Inspiration

I started with a different idea — a dialog-based app where users could choose from different AI personas to interact with. But something felt missing. I wanted to build something more practical and engaging. That’s when the idea of the AI Real Estate Agent came to me: a virtual assistant that helps people explore property listings through natural conversations and personalized AI-generated videos. Including interactive graphics and integrations such as Google Maps, with an 24/7 agent availability.

What it does

The AI Real Estate Agent is an interactive experience where users can chat with a virtual real estate agent. It responds with custom video messages (powered by Tavus) and provides real-time property data pulled from Supabase. The result is an engaging, lifelike interface for browsing and learning about real estate listings.

How I built it

• I used bolt.new to quickly build the frontend and conversation logic.
• I set up Supabase as the backend to manage and query property data.
• I integrated the Tavus API with a customised persona and system prompt to effectively handle the user requests coming from the application.
• I spent time digging into the Tavus documentation to fully understand its capabilities and come up with new ideas to enhance the experience.

Challenges I ran into

• Making the experience feel truly interactive while keeping everything smooth and responsive was a challenge.
• Communicating effectively with the Supabase backend + Tavus API took some trial and error.
• Integrating Tavus and understanding how to make the most of its API required patience and experimentation.
• Shifting gears from my original concept to a new one mid-hackathon was tough but ultimately the right move.

Accomplishments that I'm proud of

• I'm proud of how I pivoted from one idea to another and built something engaging in a short amount of time.
• I successfully integrated Tavus, Supabase, and bolt.new into one seamless experience.
• I created a working prototype that feels like it could evolve into a real product.

What I learned

• How to effectively use bolt.new with third-party services.
• How flexible the creative process can be — and how shifting direction early on can lead to a better final result.
• I was surprised by how easily Supabase integrated with bolt.new, making backend setup straightforward and fast.

What's next for AI Real Estate Agent

I'm excited to keep improving the project by:

• Expanding the property database with more detailed and varied listings, as well as adding new locations.
• Improving the AI’s ability to handle more complex user queries.
• Adding user personalization and possibly language options.
• Exploring real-world applications and potential partnerships in the real estate industry.

Built With

• bolt.new
• claude
• supabase
• tavus