Realytics

Inspiration

As new adults and near-graduates entering a challenging job market, we've faced the uncertainty of housing affordability and real estate decisions firsthand. The complexity of fluctuating markets—driven by location, economics, and global events—can make homeownership and investment feel out of reach. We realized this wasn’t just our struggle; millions are asking the same questions: "Where should I invest?" or "Can I afford to live here?"

Our project was inspired by this shared challenge. Existing tools often lack accessibility or accuracy for the average person, so we set out to build a solution that simplifies real estate data. By using two decades of historical trends and future projections, we empower users with clear, actionable insights into the market, helping them make smarter decisions.

What it does

Our project predicts real estate value for any city that the user inputs. It analyzes past real estate trends over the last two decades and generates real-time analytics and future projections for the coming years, all the way until 2035. The core features include:

Quantitative Predictions: Our tool provides exact predicted median housing prices for each selected city, alongside the current percentage change in housing value. This quantitative insight gives users a concrete understanding of the financial landscape in their chosen areas, helping them compare options effectively.
3D Heatmap: A visually dynamic heatmap that overlays the U.S. map, highlighting cities expected to have the highest growth and property value increases. It gives users an at-a-glance understanding of where opportunities lie across the country.
Trend Graphs: Users can view a detailed graph showing historical trends from the past and projected trends up until 2035. This gives users insight into how the market has evolved and where it’s likely to go, helping them make informed decisions about where and when to invest.
Real-Time Analytics: By leveraging real-time data, users can receive up-to-date information about cities of their choice, allowing them to stay on top of rapidly changing market conditions.

Our submission to the "Open Source Data" track emphasizes the power of shared data to drive innovation and solve complex, real-world problems like affordable housing and investment planning. By utilizing open-source data, we’re not just predicting market trends—we're also advocating for a future where data accessibility helps people make better, informed decisions.

How we built it

We used multiple open source datasets from Hugging Face, mostly tracking Zillow databases, as the main data source for our models. We utilized Scikit Learn, Numpy, PyTorch to train the AI prediction models and perform testing and evaluation. For the backend, we used Flask and SQLite to create API endpoints to the model, AI-powered suggestions, and the AI assistant. For the frontend, we used Next.js and Tailwind CSS. Lastly, we used AWS to host and deploy.

Challenges we ran into

Because of the numerous number of API endpoints, we had challenges during the hosting and deployment process, getting all of it to run live. We also worked on the deployment for almost six hours, and we're so proud it worked finally!!!!

Accomplishments that we're proud of

We will able to create a model that predicts real estate prices pretty well as well as with a very interactive user interface with helpful suggestions and assistant features. The visuals and analytics are easy to understand and show very detailed projections.

What we learned

We learned a lot about training and finetuning a model pretty much from the ground up using open source databases, as well as a lot about hosting our own backend through AWS. Each member originally had pretty different strengths, and though we played to them, our collaboration allowed us to learn a lot of skills in areas we had less exposure in before.

What's next for Realytics

Compare two cities function so that users can compare the analytics for two different regions, as well as adding even more regions/cities to the map for users to explore!