Shopify F22 Data Engineer Intern Challenge

The following code base implements the image repository from the Shopify Data Engineer Intern document.

Live Demo (might be down) : link

Instructions

The project has been containerized. Just run the following commands:

docker build -t image_repo:latest .
docker run -p 5000:5000 image_repo

Then visit http://localhost:5000/ on your browser to test it out.

What was done

• I chose to fully implement the SEARCH functionality. More on this later.

• I have also partly implemeted the ADD functionality for testing purposes. No bulk upload/secure storage features were implemented.

What's cool about it?

1. Searching images by similarity was an open-ended design choice. I chose not to use ML libraries due to added complexity. Rather, I computed perceptual image hashes to compare image similarity. It's not perfect, but it's simple and performant.
2. I tried to extract tags both from the user input and the filename. That should allow for more accurate tag searching.

How does it work?

Homepage

Search

Now you have two options to search images by:

By tag

Searching by 'car' as an example... Result:

By similarity

Similarly, one can search by similarity. Upload a photo to find corresponding similar images in the repository.

Upload

Successful uploads will show up as the following:

Dev Documentation

The program has been written in Flask and uses a simple MVC model. There are separate directories for the models, views (templates) and the controllers.

Most functions have been well documented in code using comments. Due to the small nature of the program, the code base is self-documenting.

Things I skipped

Due to time constaints, I couldn't add test cases and had to rely on testing the app through the UI.