RealRisk

Map and Chatbot
Risk Scores
Extra Information

Inspiration

The development of housing communities depends on multiple elements, one being the surrounding environment in which it is being developed. It's crucial for home builders to conduct through analysis of the surrounding analysis. These risk factors greatly affect the success of a housing community if it is built. We were interested in this challenge from Lennar due to its practical application and real-world links. We also saw the potential of AI to be used to improve the calculation and generate human readable insights.

What it does

RealRisk allows users to input an address, including the street name, city, state, and zip code, or the coordinates to receive a detailed report consisting of scores for demographics, competitors and market, environmental risks, regulatory risks, and crime statistics. We found these categories to be important for home builders before making an investment in the area to ensure that profitability is maximized.

How we built it

We used multiple data sources including the U.S. Census, The American Community Survey, Yelp, and we enabled our LLM implementation to search Google to get the most updated results to power our calculations. We found this to be necessary because of the variety between implementations of regulations and the availability of data nationwide.

Challenges we ran into

We were attempting to use APN for generating the insights, however, many databases were pay-walled or not publicly available. We also found it difficult to find a universal solution. Therefore, the data would have been inconsistent and incorrect. We hope to be able to use APN in the future when we get more time and funding to accurately poll the data.

We also ran into problems dealing with addresses and geocoding. There was times where it was very difficult to work with geographic coordinates especially in rural areas because we had to determine what the zip code and county they were in.

Accomplishments that we're proud of

We integrated Google Maps into our app to allow users to have a better visuals to locate and recognize addresses. We also were able to get up-to-date information by using our LLM implementation using GPT-4o to search Google to get the most up to date results.

Additionally, our front-end has various elements to it, including not only the risk scores and their assessments, but more information about the area, such as land area and points of interest in the surrounding communities.

Lastly, we integrated GPT-4o for our chatbot feature in our app. This chatbot enables users to ask questions regarding the information provided by the app and provides an in-depth explanation to user's questions.

What we learned

We developed a deeper understanding of Python, AI technologies, how to integrate front-end and back-end cohesively, and how to integrate APIs and large amounts of data into our code.

What's next for RealRisk

In the future, we would like to implement APN to better reflect industry standards. We would also like to have access to more datasets to better train the LLMs in order to have an even more accurate response for users when it comes to generating human readable insights. Given more time, we would like to integrate more data points to better inform users on the risk of investing in certain areas. This is very easy to do because of our robust backend implemenation.