Watt-son

Make it easier for people to explore solar for their house (GPT came up with this name because Watt: Power and Watson: because GPT is being investigative here)

Comment

I am in-person so I did not make a video, but the form wont let me submit the project without adding a URL, so please check out my video as well.

Inspiration

  • There's billion dollar companies who do not provide the solar feasibility data to a homeowner unless the customer signs up for an 'onboarding call' with the sales team of that company. It should be easier to find out this data and there's no need to have to talk to a sales consultant.

What it does

  • All this needs is a user's address and their average monthly electricity bill. Using this, it can identify different contract types for a user to go solar, how much they could save in $$ and emissions, and many more.

How we built it

  • I used GPTs with the ability to retrieve incentive info from embeddings + solar API endpoints provided by google to retrieve data. The inputs to these functions are done through OpenAI function calling where it retrieves relevant information from natural language by understanding customer intent.

Challenges we ran into

  • Debugging on the GPT builder + writing the spec. I never wrote a spec in the past, so figuring how to write that and making it work (I partially made it work, still working on the latter half :/. On the upside, I think I'm an hour or two away from figuring it [famous last words] )

Accomplishments that we're proud of

  • Getting this done in a very limited amount of time. I only got access to GPTs today morning and could not find much time because I had to also work, but no complaints, this was extremely fun.

What we learned

AI is

What's next for Wattson

  • Building an assistant using OpenAI API docs, building a website on top of it
  • Rendering graphics of different layers of solar generation (how would panels look on their roof, which parts of their roof are better for solar, which parts of roof receive how much solar per year) and letting the user interact more with the API
  • Explore contract options, the right size of solar for their use case, and understand how much they can save, all through a simple interface.

(I started writing this 15 minutes before the deadline so you can notice it was rushed up)

Built With

Share this project:

Updates