Grokleads

Inspiration

Parsa Tajik

What it does

It takes datasets of hooks (notifications) on website prompting customer to interact with chatbot and uses sparse autoencoder features to find correlations in the data and what improves click rates. Then, users can iterate and change the system prompt that generates the hook

Once the user clicks the widget, the chatbot uses x search, collections (product catalog), and web search to find latest trends, convincing the user to buy a shirt

How we built it

Sleeping 1 hour

Challenges we ran into

xAI API streaming, integration hell, converting dataset formats from parquet to csv etc, database schemas xAI typescript sdk with vercel is kinda weird for tools? Only works with python sdk

Accomplishments that we're proud of

It sorta works?

What we learned

Streaming is hard, features are cool

What's next for Grokleads

Make a startup

Built With

• nextjs
• postgresql
• python
• x
• x-api
• xai