-
Landing page
-
Enter the main query
-
Final results
Inspiration People spend a lot of time online shopping, scrolling through hundreds of items after typing simple keywords like “blue dress.” But in real life, we shop based on more than just color — we think about fabric, fit, price, where we want to wear it, and even things like return policies. I wanted to create something that understands all of that and helps users find exactly what they’re looking for, without wasting time.
What it does Perfecta.ai – The Perfect Pick is a smart shopping assistant that talks to the user like a real shopping expert. It asks what you like — your preferred style, material, color, budget, occasion, and more — and then searches a product list to show the best matches. It makes online shopping feel more personal, just like how you’d get help in a real store.
How does it differ Perfecta.ai remembers all your likes and dislikes — from fabric and fit to style and occasion. Unlike basic filters, it has a real conversation with you and adjusts based on your feedback. It works like a real personal stylist, helping you find the perfect pick every time.
How we built it I used React to build the chatbot-style frontend that collects the user’s preferences step by step.
The backend is a agent built using Toolhouse that matches these preferences to a small sample product dataset on web.
Each product has extra details like reviews, return policies, shipping info, and prices.
Everything is designed in a way that can later connect with online shopping websites, voice tools, or extensions.
Challenges we ran into Designing the flow of questions without making it feel too long or boring.
Matching products properly with user input using simple logic and a small dataset.
Making the chatbot feel friendly, helpful, and not too robotic.
Accomplishments that we're proud of A working prototype that can take inputs and give product suggestions in real time.
Built the base for something that can grow into a browser extension or voice tool.
Created something that feels more natural than normal filters or search boxes.
People who tried it found it fun and useful.
What we learned It’s tricky to make AI feel human, but users love it when you get it right.
Structured data is very important for good results.
Simpler tools can work really well if they are planned properly.
The more personal the experience feels, the more helpful it is.
What’s next for Perfecta.ai Add support for more natural, free-form inputs like “I want something classy but under $50.”
Connect to real shopping sites or APIs to search live products.
Build a browser extension version that works while people are already shopping.
Try voice-based shopping where users can just talk and get results.
Partner with fashion or lifestyle brands to offer it on their platforms.
Log in or sign up for Devpost to join the conversation.