Python Task Submission: Akash Chattopadhyay

[antremis]

This is the submission for the Python Task for Urban Match.

I have tried to improve the directory structure for maintainability on a bigger code base.

The approach for the recommendations used is on the following criteria:

1. Users should be from the same cities
2. Users should both be below 18 or both be above 18.
3. There can be up to 2 years of gap in age between users.
4. Users should have a non-zero overlap in interests.
5. Recommendations should be sorted by the number of common interests in increasing order

Assuming a large user base, I have attempted to do the recommendation logic on the DB to offload responsibility in interest of throughput.
A better solution would be to perform a pre-computed cosine similarity on all unique interests and compute an interest-interest cosine similarity matrix. A user can then be recommended individuals based on their cosine similarities. However, I have used a more basic approach in the interest of time.