Inspiration
All three of us experienced cases where lead generation would have been super helpful. For example, finding startups for venture capital firms, PPE during a pandemic for healthcare facilities/universities, and suppliers for university ITS departments operate safely and efficiently.
What it does
Sheetz is an AI engine for converting desires to data in a sheet. For example, if you wanted to get the phone number, location, name, and website link for hundreds of suppliers for PPE, you could simply ask Sheetz, which would produce a sheet that has relevant information (from the Internet) ordered in the user's desired structure.
How we built it
We built Sheetz using NextJS, OpenAI, Exa, MongoDB, Express / NodeJS, TS, Vercel, and Firebase.
Challenges we ran into
- Individually updating cells on the user's end when the data was found
- Optimizing latency by using parallel processing to execute multiple calls at once
- Using BFS on webpages to find the answer to a query not on the current page
Accomplishments that we're proud of
- Frontend Freaks (aka Clean UI)
- Adding a depth field to parse relevant pages to answer a specific query if the answer is not on the initial web page
- Creating prompts to abstract the data generation process from users
- Converting aggregated data to a downloadable format for user
What we learned
- The constraints on when to stop/continue BFS through related web pages
- The accuracy limitations of parsing web scraped text using LLMs
What's next for Sheetz
- Integration with Google Sheets API for direct export to Google Drive
- Editing/selective querying for specific cells on a Sheetz
- Optimizations to limit monetary costs for collecting large amounts of information
Built With
- exa
- firebase
- mongodb
- next.js
- node.js
- openai
- typescript
- vercel
Log in or sign up for Devpost to join the conversation.