New Horizons

New Horizons

Inspiration

As college students, we are always looking toward the future. We dream about buying a home and also about fighting climate change. We also really enjoy using the convenience of ChatGPT and other AI tools. Accessing helpful, relevant information only using a one-sentence prompt has become necessary. We realized that the search for a house that fits all of a home buyer's needs could benefit from this "one prompt" mentality. If we could streamline the home search process to a single search prompt, house searching would be a more effortless, more enjoyable experience. Often, buying a house is one of the most critical decisions of an adult's life, so New Horizons was born to reduce the anxiety of purchasing a home and as a new tool for buyers and real estate agents to connect.

What it does

New Horizons first takes a search query from the user. The user may type the qualities of houses they want or any qualitative information about them. It uses state-of-the-art machine learning to find the best match for the users. It displays relevant information about the house with images and a panorama view, allowing you to schedule an appointment with the realtor.

How we built it

We built the program using React and Flask to integrate machine learning seamlessly into the backend. Our choice of language consists of typescript, CSS, HTML, and Python. This is a list of powerful libraries we used:

React: Frontend library that enables us to develop web apps using typescript, CSS, and HTML.
Flasks: The backend library lets us connect to our front end while performing data-intensive tasks.
Sentence-Transformers: Convert a word with context meaning into a high-dimensional vector.
OpenAI API: Chatbot that takes in fragments of detailed descriptions and creates concise summaries.
Numpy: Allows us to do advanced math such as linear algebra, vector calculus, and probability.
Pandas: The statistical library allows us to read CSV files and data science.

Challenges we ran into

Frontend and backend communication: Communicating seamlessly in real-time with front and back was challenging.
Not enough data for our machine learning algorithm: The Messlia-free dataset was not vast, and there was no exact data set on housing descriptions; we had to input datasets manually.
Integrating both frontend, backend, and machine learning: Integrating all the different code pieces wasn't easy.
OpenAI Update: OpenAI recently updated their SDK for ChatGPT API, thus requiring us to read through their syntax update.
Limited hardware: Only one person had GPU on their laptop, and the rest of the users were required to use Google Colab.

Accomplishments that we're proud of

First timers: Three of four were first-time hackers with limited coding experience.
Both front and back end integration: Seamlessly update with Server Side Events(SSE) and Flask Cor in real-time.
Machine learning: Applying machine learning algorithms from class.

What we learned

Many of us have not worked in a group setting on a project with limited time. We learned the difficulties of project management and workflow. We learned more about using React and Flask and solving complex problems that initially seemed easy. We learned how to build a virtual environment using Google Colab or Anaconda. Most importantly, we persevered through hardship when we received errors.

What's next for New Horizons

Moving forward, we plan to:

Develop a React Native and Swift: Having Android and IOS apps would significantly improve the quality of life for the end users.
Integrated with prominent real estate companies: Access to extensive data would improve the quality of our product.
Starting a Startup: