Inspiration The idea for this project came directly from a personal experience. When I was looking to rent out a house, I realized I actually did not know much about the surrounding area. Even after purchasing the property, the return on investment (ROI) did not increase much over the years. We were lucky not to run into any major issues, but the entire process felt like a blind guess. I realized I wanted an application that could give a complete, transparent picture of a place before making a decision. I needed something that clearly showed a location's accessibility, available services, safety records, past incidents, historical ROI trends, and local facilities all in one place.
What it does The application acts as an automated location analyzer. When a user selects a specific location on the map, the system immediately evaluates the existing conditions of that area. It breaks down the location's real-world data into clear categories like economic trends, environmental factors, transit accessibility, and safety and provides a straightforward, easy-to-read informational report about what it is actually like to live or invest there.
How we built it We developed the core application using Python to handle the backend logic and data routing. The intelligence behind the analysis is powered by the Gemini API. Rather than relying on a single, general AI prompt, we designed the system to use Gemini as distinct, specialized agents. We have specific agents handling different domains: one for infrastructure and ROI, one for safety, and one for transit. When a user drops a pin, our code captures the latitude and longitude and pulls local geospatial data for that exact spot. To ensure the AI grounds its analysis in reality, we calculate the exact physical proximity to local amenities. For example, the underlying logic uses standard distance calculations to measure how far a property is from public services or transit lines. This forces the AI to base its reports on actual geographic facts rather than making assumptions.
Challenges we faced Building this was not straightforward. We ran into several persistent issues in the code, especially when trying to integrate various data sources and coordinate systems reliably. The biggest challenge, however, was managing the AI agents. Initially, the agents would sometimes overlap in their analysis, provide conflicting information, or misinterpret the raw data we fed them. We had to go back and iterate on the system multiple times. We spent a significant amount of time testing and refining the prompts to force the agents to be accurate, concise, and strictly analytical.
What we learned We learned a massive amount about applied AI, specifically how to constrain a large language model to act as a strict data analyst rather than a conversational chatbot. We also gained practical experience handling geographic data, working with spatial math in code, and debugging complex API integrations under the pressure of a hackathon.
Log in or sign up for Devpost to join the conversation.