Inspiration
I was inspired by watching professional real estate agents research and evaluate comparable properties using sources like the MLS, Google, and county appraiser data. I wanted to create an AI agent to help with this process.
What it does
My project is a "Real Estate AI Agent" that assists real estate agents by finding comparable properties for a specific listing. It uses data from MLS, Google, and county appraisers to generate a report based on professional realtor criteria.
How we built it
I built it with Agent Development Kit (ADK) using a SequentialAgent and then several SubAgents: Google Maps via MCP, Google Search using the built-in tool, MLS listings via function tools, and a Report Writer Agent
Challenges we ran into
The main challenges were translating subjective realtor criteria into effective prompts and integrating data from various sources. Additionally, as an experienced programmer, it was a significant challenge to force myself to guide the AI using prompts in English rather than directly coding the solution. This required a shift in mindset and approach.
Accomplishments that we're proud of
I'm proud that it's a useful tool for my real estate colleagues. They've given positive feedback on the accuracy and efficiency of the comparable property reports.
What we learned
I learned the importance of clear and iterative prompt design when working with LLMs. It pushed me to use language to guide the AI instead of just writing code.
Also I loved the google_search tool to extract information from the Google Index; to retrieve non-structured information about the property.
What's next for the project
I plan to expand it into a toolkit with more features for US realtors, including market analysis, property valuation, and lead generation.
Log in or sign up for Devpost to join the conversation.