AskRea

• 

  Plug-in logo

• 

  Plug-in output

• 

  Excel model

Inspiration

We were inspired by our experience in the real estate industry. Despite having access to millions of data points, because of time constraints, the average real estate professional still spends a lot of their time making data-poor “gut checks” on what deals to pursue.

Because of the volume of deal flow, these "gut checks" can take up to 30% of their time and still often lead to missing potential great investments.

What it does

AskRea solves this by acting as the analyst that every real estate professional wishes they had. Ask Rea to analyze any investment deal, and she will immediately tell you what the potential economics are, spin-up a basic pro-forma, and most importantly, tell you whether the deal is worth a more detailed look.

How we built it

We built our project by separating it into 4 simple steps:

Get property information: After a user submits a Zillow link, AskRea retrieves all relevant information about the property from the ZillowAPI, including sales price, zoning information, location, and size.

Get zoning information: After receiving the zoning information from Zillow, AskRea queries the local city zoning code (in this case, Sacramento) to understand the zoning & building considerations. We use a retrieval-based LLM approach here (through LangChain) to prevent hallucination.

Get comps: After getting the zoning considerations, AskRea creates an estimate of what can be built on the property (eg 1-bedroom home, duplex) and searches the Zillow API for similar properties nearby to get comps and estimate the potential final property sales price.

Create analysis: Finally, AskRea generates a pro-forma model (based on an existing template) and conducts a financial analysis. From this analysis, she suggests whether the property is worth a second look. The pro forma (in Google Sheets) was created through a Zapier WebHook, and edited directly with the Google API.

Challenges we ran into

We were struggling to figure out a solution to understanding the building and zoning considerations of the project. Standard GPT-4 was getting the answer right 80% of the time, but did not have the specific local knowledge to always generate a correct answer. That's when we decided to query specifically the city zoning code using LangChain.

Accomplishments that we're proud of

We're most proud of creating an end-to-end plug-in flow that could be useful to developers today! Stitching together all the disparate pieces was a big accomplishment, and we created a product that thousands of people could use in 24 hours.

What we learned

We learned that of the entire process, the actual creation of the plug-in itself was probably the easiest and most straightforward part. It's given us both confidence to create more plug-ins in the future.

What's next for AskRea

AskRea has the potential to be a full-fledged real estate analyst. We started with creating feasibility analyses, but AskRea can add value throughout the entire development process- creating detailed financial models, performing cost estimates, generate optimal designs based on financial and zoning consideration, and update plans based on new inputs. We plan to build-out these capabilities next. ....

Built With

• chroma
• langchain
• openai
• python
• sheets