AmpliPoodle

Inspiration

We were initially intrigued by the idea of using AI-generated images to visualize data to help people process and understand information easily. We wanted to investigate the role that images could play in understanding data and helping people make the right decision based on the data they see.

We started with using dog datasets because who doesn’t love dogs?! We were inspired by Ben and his gorgeous pup Casper, and his tracking device. It occurred to us that there could potentially be lots of data about dogs and pets available that can help their owners understand their pets better. The idea somehow evolved into looking at the different categories they could go under.

What it does

AmpliPoodle is a data visualization tool that uses AI-generated images to help us have a better understanding of our personas.

How we built it

Data: Generated activity data for different dog breeds from dog population, breed activity and percentile distribution of dog breeds in the UK.

Amplitude mapping: Data generated by faking events based on datasets. Using a set of 3000 dogs and sending events for their activity over a month. Use this to create cohorts and personas on Amplitude.

Destination: Create an amplitude destination (Configuration might just be an API key and an URL) to send data to a server that generates an image on unique URL(Hardcoded for now).

Server: Webhook that can receive Amplitude Persona/Cohort information and generate a single phrase that will be used to generate an AI image. Endpoint that generates an image and sends an image based on most recent phrase generated.

Dashboard Image

Challenges we ran into

With constant changes to our idea, we went from Crime stats to Dogs :)

What we learned

We learned how data clustering works on Amplitutde.

What’s next for AmpliPoodle

AmpliPoodle is a concept built on how we can use AI image generation to create visualization that businesses can use for their Buyer Personas. Plugging in different types of user properties and using more specific data values.