StepBack

Inspiration

Upon arrival at Boilermake we attended the Blip Workshop where we met Antonio who talked about Blip and what they do. What particularly caught our eye were some ideas that he gave out for potential hackathon projects. The idea of automating A/B testing using AI really stood out to us and we decided to base our submission around it.

What it does

StepBack uses AI-generated personas to simulate a shopping experience on a client’s website. Clients can upload images or videos of their products, and the personas provide feedback on which products they would purchase, which they wouldn’t, and why. This information is critical to business owners for improving their profits while completely avoiding the costs associated with A/B testing.

How we built it

On the backend, we used Gemini to generate the personas. We then created the frontend, allowing clients to upload product images or videos. These inputs are processed using Gemini’s API, and the results are displayed to the client in an organized CSV table.

Challenges we ran into

One challenge we faced was integrating different code segments, which we overcame using GitHub and Google Colab. Formatting and displaying frontend results was another hurdle. Additionally, selecting the right AI model and understanding its limitations required significant consideration.

Accomplishments that we're proud of

We successfully integrated AI with multiple components into a cohesive application. StepBack smoothly transitions from image or video inputs to an organized CSV summary, providing valuable insights for clients.

What we learned

This project was a major learning experience. We explored autonomous computer control, worked with APIs like Gemini, and gained frontend development skills. Our conversation with Antonio also deepened our appreciation for real-world applications of our project.

What's next for StepBack

Moving forward, we aim to enhance StepBack by incorporating video-based interactions where personas can navigate websites for more realistic shopping experiences. We also plan to introduce personalized product recommendations based on persona behavior.