Online shopping is bigger than ever before, and there's so much untapped potential. Kata brings users content they will love. How? They follow curators of content that they actually care about.

For example, in the huge market of online fashion, it can be difficult for consumers to find products that they like across so many retailers and websites. Instagram, Pinterest, Tumblr, etc. make it easy to find people with great senses of style. Now, those who know where to find the best products can share them with their followers on Kata. With a filtering system, users can also ensure they don't see content that is out of stock, not in their size, or doesn't ship to their country,

What it does

Kata provides a platform for curators to recommend products to their followers. In addition, there are features to filter out products that are not available to the user and find similar items through computer vision/machine learning.

On the backend, we are able to aggregate all the data that we're getting from these users and use it to further drive our product recommendation system.

There is another component to Kata: The Google Assistant app. Here, users will be able to access our product recommendation system directly, and be able to get suggestions regardless of what other users are posting on their feeds. This input helps feed in more data to further train our product recommendation system.

How we built it

We built the frontend in AngularJS, the backend in Flask, and the Google Home app in Nodejs' actions-on-google. We also utilized AWS RDS and Lambda to deploy and store all the data that we've stored. We were also able to tunnel a lot of traffic to our custom endpoints using ngrok.

We also implemented a lot of machine learning elements in our frontend website. We created a search feature that allows you to search through existing elements using a link to an image. This works by classifying the image and then querying the database for keywords generated from the image. In addition, our product recommendation system runs on a model that continually generates connections between two objects should they become associated with each other, so that our recommendation system will slowly start to associate more and more objects in order to provide a better recommendation system.

Challenges we ran into

We ran into a lot of issues with AWS RDS not allowing external users, so we made a really hacky workaround by tunneling connections to my computer in order to commit to the database. AWS really needs to fix this by making it more intuitive.

Accomplishments that we're proud of

We're proud of our team for managing to push through a lot of really, really dumb bugs.

What we learned

NodeJS is truly great. AWS has issues with VPS groups. It sucks, but it's not the worst.

What's next for Kata

We plan on building this out! We already have a stable revenue stream through our recommendation system (every time a user adds one of our items from our list, we make a cut). We will focus on our social network platform, getting more curators and users in order to entice advertisers to advertise on our platform. We're better than other traditional social network platforms in that we offer more targeted advertising. We know who you follow, so we know pretty much what you're interested in.

Share this project: