RealVest

The results and summary on our website, realvest.ai

Inspiration

The world of commercial real estate can be an intimidating venture for individual investors. The intricate process of property sourcing, conducting thorough research, and deciphering complex financial analyses often presents a barrier, creating a lack of transparency and accessibility.

What it does

RealVest seamlessly integrates real-time commercial real estate listings and pertinent data sources. This simplifies the process for investors by tailoring the search experience according to their unique circumstances, enabling them to efficiently navigate through the myriad of property listings.

How we built it

We've constructed a real-time data pipeline that aggregates information from top-tier real estate websites. The data is then housed in our proprietary databases, both relational (Postgres) and vector-based (PineCone). Utilizing these resources, we've built a semantic search and ranking system to streamline the search process. Further enhancing the user experience, we utilize Language Models, specifically OpenAI GPT and Cohere Command, to summarize and present the search results in a user-friendly, natural language format.

Challenges we ran into

Parsing and assimilating structured data from diverse sources proved to be a substantial challenge due to the variability in data formats and quality. Fine-tuning the search ranking algorithm to optimize relevance was another complex task. Additionally, we faced time constraints which necessitated the narrowing down of our data sources, making data selection a critical task.

Accomplishments that we're proud of

We take immense pride in successfully completing a robust data ingestion pipeline that feeds into our unique blend of relational and vector databases. Our end-to-end solution, while currently constrained in scope, is fully operational and capable of conducting searches, presenting results that are both useful and relevant to the user. This achievement validates our approach and fuels our motivation to continuously refine and expand our platform.

What we learned

Our journey allowed us to gain practical experience in utilizing embeddings and vector databases for hybrid searches, a novel approach in the realm of property searches. We've also gained insight into the challenges inherent to this sector, acquiring a clearer understanding of the tasks that we need to address moving forward. This experience has enhanced our perspective and prepared us for the roadmap ahead.

What's next for RealVest

Our roadmap includes further refinement of our semantic search capabilities and development of a robust recommendation service. We plan to partner with a wider array of data providers to broaden our data coverage. As an immediate next step, we will release a market prototype to glean invaluable feedback from early adopters. This iterative process will guide us in continuously improving our platform to better serve our users.