Chrys

With rising interest rates and the current housing market, young homeowners must be informed of the financial responsibility of purchasing and maintaining a home. Luckily, Chrys is here to guide them!

Comment

Inspiration

Recently, one of our team members has been involved in refinancing her parents' home mortgage. Looking at how complicated the process is, we were all motivated to develop a web app targeting young homeowners learning to navigate the financial aspects of an important milestone. This is the kind of tool we would love to have as we approach college graduation and build lives of our own!

What it does

This web app utilizes an AI chatbot to prompt users for information about their goals like buying a new home or refinancing their current properties. In return, they will get potential mortgage rates, insurance options, local property tax information, and more.

How we built it

We used a variety of APIs, libraries, and tools. Some APIs we utilized include SambaNova (hosting a Llama 3.1 LLM), MongoDB Atlas (storing user data), and Mortgage Rate (calling current mortgage rate information). In addition, Python's Streamlit library was useful in designing and hosting our website. Meanwhile, Hashicorp Terraform helped speed up the database configuration process and improved code expandability and maintainability as a powerful infrastructure as code tool.

Challenges we ran into

Framework Decision: Since the majority of our members were unfamiliar with LLMs prior to this project, choosing the right multi-agent framework for the application's needs was a challenge from the start. Even after deciding on a framework, configuring the agents in a way that required minimal prompt engineering was a lengthy task.

Database Input: Although most of our team is familiar with MongoDB, the unique goals of our application made it harder to store the desired information in the database. Since a major highlight of this tool is to make it easily accessible and useable for beginners, we wanted to skip common questionnaires that many similar services require and only prompt the user for data (about their down payments, salary, credit score, etc.) when necessary. As a result, we rely on our LLM to find and store desired data in a user's response.

User Experience & Interaction: From the beginning, we were set on designing an intuitive and user-friendly UI for young homeowners unfamiliar with mortgage terminologies, but our team is still a group of college students with minimal knowledge about the topic. Therefore, we ended up needing to devote a lot of time to research to ensure that our chatbot's responses were accurate and helpful.

Accomplishments that we're proud of

Efficient Setup: Successfully setting up MongoDB Atlas with Terraform on the first try--Go Nidhi!

AI Model Deployment: Hosting Llama 3.1 on SambaNova cloud and integrating it effectively into the chatbot.

UI Development: Seamlessly hosting a multipage Streamlit website offering users an intuitive interaction experience.

User Data Retrieval: Due to our diligent efforts, we were able to handle user data in a low-friction manner.

What we learned

Teamwork: Leveraging individual strengths, such as Terraform expertise, accelerated the setup process.

Mortgage Basics: Gaining insights into the financial hurdles young buyers face, including terms like principal, and interest rate trends.

Cloud Hosting: Effectively using SambaNova's cloud services to power AI solutions and MongoDB Atlas to store user data on the cloud.

What's next for Chrys

Enhanced Features: Adding personalized financial advice based on the user's profile, including saving plans and home affordability calculators.

Integration: Incorporating additional APIs for broader financial data, such as insurance comparisons, Zillow API for local home prices, and more.

Built With

Share this project:

Updates