Inspiration
We were inspired by our personal experiences with first-time homebuying and felt pulled to make an impact in this area of the real estate industry.
What it does
Our web application leverages agentic technologies to create a multi-purpose virtual real estate agent, specifically targeting first-time homebuyers who are new to the process. What our agent does is take in user responses, store and update its memory context regarding its different users, and then use that memory in the future to better serve our users. HomeBound is a one-stop shop for your search for a new home. It helps you filter out which property options would best serve you and your needs.
How we built it
We split this project into three main components: frontend, backend, and our AI agent. Our LLM of choice, Anthropic, powers our agent, which acts as a link between the backend logic and the frontend interface. We built out the backend using Python, PyTorch, Flask, and SQLite, then connected it to our agent. We then set up our frontend using the React framework as well as JavaScript and integrated the three components together to create a streamlined data flow, powering our core features for HomeBound.
Challenges we ran into
Our main challenges fell into two categories: backend personalization and frontend integration. Our main challenges fell into two categories: backend personalization and frontend integration. When building out the backend, we found that creating a memory system that both accurately scores the different properties while taking into account the evolving preferences of the user was a challenging and iterative process. We adjusted our scoring system throughout development in response to our testing process, ensuring that our users would receive more personalized service. Other issues we ran into in the process of integrating our frontend included data persistence when it came to React components and features like chat history or memory logs. It was important that data was consistent across user sessions, so validating and testing different use cases and edge cases was a priority and took up the bulk of our time.
What we're proud
We’re proud that we identified a real-world pain point in our communities and built a viable solution that can streamline the homebuying process and reduce unnecessary stress for first-time homebuyers.
Log in or sign up for Devpost to join the conversation.