Trendswipe

Inspiration

Decisions can be tricky - they take up a lot of your energy and consume an enormous amount of time. We feel that in so many aspects of our lives as University students. We want to feel the best we can and adopt a style that makes us feel the best version of us! We decided to make discovering products as effortless and fun as swiping through social media, so we built Trendswipe: a Tinder-style shopping experience that learns your unique style with every swipe.

What it does

Trendswipe transforms online shopping into an addictive, personalized discovery game. Users swipe right on products they love and left on what they don’t - just like Tinder. Behind the scenes, our AI analyzes swipe patterns in real-time to build a dynamic taste profile, serving up increasingly relevant recommendations from both fashion and beauty catalogues. Users can save favorites, view matched products, and get AI-powered style advice through an integrated chatbot. Shopping becomes entertainment, not a chore—fast, fun, and perfectly tailored to you.

How we built it

We decided to tailor Trendswipe as progressive mobile oriented web app to get the best out of the swipe feature, our tech stack was:

• React: Smooth, gesture based swipe interface that feels natural
• Flask: Handling user sessions, product recommendations and AI model interface
• Supabase: for simple and real-time database management and authentication. Storing user preferences and swipe history.
• Gemini AI: Integrated for personalised style recommendations and conversational shopping assistance
• Machine Learning: Algorithms that process user interactions to refine product matching based on price sensitivity, category preferences, and visual style
• Grafana: for a comprehensive overview of all the key statistics and trends like user engagement and number of products and price

Challenges we ran into

Mobile responsiveness was very challenging to implement, the swipe gestures took a long time to tailor and get just right for a balance of natural force across the screen and preventing lag from long animations. We tried to implement agents but spend lots of hours troubleshooting the agent class which was very difficult to embed.

Accomplishments that we're proud of

Our frontend was implemented brilliantly and looks amazing as well as work seamlessly, the swipe feature functions very well. Our AI matching algorithm that learns the users product specifications and personalises the next recommendations based on their previous swipe history. The Grafana dashboard summarises many core user and product metrics well and gives an excellent overview of trends and current user habits.

What we learned

Mobile-first design principles was a big achievement for us, it is used as a standard in industry and we really tried to implement it in a professional and coordinated style. Grafana was a big step up for some of us as we have never really used it before, so starting from scratch was difficult but it was rewarding because we learned how to make a really engaging and beautiful dashboard. We learned that implementing agentic AI functions are very difficult and possibly to learn to integrate it better in our own personal projects in the future before diving head first into something as difficult as that.

What's next for Trendswipe

Smarter recommendations, we plan on using a deep neural network to capture nuanced style preferences and predict trends more accurately. More game features, already expanding on our 'fashion odyssey' game progression to fully engage our users in a meaningful and fun way. A full retailer integration would be ideal as we don't have many actual retail options, we want to expand on our designer brands and upload products we know our users will love. A more analytical Grafana dashboard would be amazing for looking at more user trends and use that information to further strengthen our AI models.

Built With

• flask
• gemini
• grafana
• postgresql
• react
• supabase