Catalyzers.

recommendation page 2
profile edit page
recommendation page 1
home page - dashboard

The Problem Identified:

Amazon wanted a better environment and a sustainable solution for everyone to add to their system. Something that benefits the environment, society and individuals. We identified that, RETURNS on an every online shipment is trifold the loss of money made plus an additional burdent on environment.

The Solution:

Most RETURNS [53/-] in clothing market [25/- of net orders] are for a sole reason - {size/ fit/ comfort issue} - if just by adding an intelligent layer in between which can give me personalized recommendation according to my buying patterns/ reviews of other users and product details... A confidence score of the product, which makes me sure of the fit and not having the need to return it just for a mere size issue. This sole thing could save the majority of a sustainibility issue and can be a win-win for everyone.

Inspiration

Our biggest inspiration to have this were the problems identified by people around us as we discussed. We talked to people, understood their concerns, most of them had the issue of size while ordering online and they could do nothing of that except ordering and returning again... We, not only found a solution, but also, our first set of users who tried our solution live.

What it does

It analyses your purchasing patterns, orders you had in the past, your body fit and measurements, with the reviews of product you are looking at and it's size + fit description. Based on this, it gives a confidence scoring to each product of how likely it would be a good fit to the user.

How we built it

We built a web-widget for the solution, our tech-stack being javascript for backend, html and CSS for frontend. we did code with assistance of AI.

Challenges we ran into

Our main challanges is to keep the widget assist us while live shopping, while doing the job of building a personalized recommendation and keeping it fast with token effecient. What we found best was to use an hybrid model of algorithms and mathematical engine to load most part of the scoring under 0.1 seconds and then do the api call to have a justification statement with gemini 2.5 flash api key to give the personalized recommendation. We managed to keep the thing super fast with maintaining lowest possible api key usage.

Accomplishments that we're proud of

The pace at which we developed a working solution. The ideas we brought to have the mathematical engine with the ai model to have both speed and accuracy. The solution which we expect to solve can reduce upto 30 percent of returns which is a big WIN-WIN for the industry, sellers, customer, environment and Amazon.

What we learned

We learnt that the essence of problem solving lies in having the minimum change which causes the maximum impact. Hence, our solution being a simple, ready to go extension.