At Udemy, I suggested and built around half of the most successful features that were built during my time there. Products I suggested and built include: Course Feed, Notifications, Wishlist and Course Previews.
I also built a new course landing page -the most complex page of the website at the time- from scratch, built note taking functionality into the course taking experience and worked on many other projects big and small.
Initiated and finalized infrastructure projects like Memcache integration, Redis-based backend for notifications and course feeds. Decreased the API response time by 50% by heavily profiling it and optimizing the codebase. Initiated some engineering processes for the organization, i.e, carried over the codebase from SVN to Git and started the code review process.
Position: Software Engineer
San Francisco
Note: Some of the products below went through visual redesigns after I implemented them. Unfortunately I don't have the screenshots of those from back when I implemented them.
Correct way of sorting the reviews:
Score = Lower bound of Wilson score confidence interval for a Bernoulli parameter
We need to balance the proportion of positive ratings with the uncertainty of a small number of observations.
What we want to ask is: Given the ratings I have, there is a 95% chance that the “real” fraction of positive ratings is at least what?
Implemented the 'course-box'
Came up with this idea and implemented with the CEO's support though most of the company thought this would backlash since people would just download all the content quickly.
Came up with the idea and implemented feeds: one for each user(which we ended up retiring) and one for each course(which ended up one of the most successful products we have ever launched), so, people can easily follow the activity happening in a course and also easily see what other interesting users and their friends are up to.
Implemented these feeds with personalization(persistent aggregations for individual users) when we had 100K users and the system was still in use when Udemy reached 4M users without any major modifications.
One of the most challenging projects built at Udemy. Notifications are personalized, aggregated notifications need to be persisted and one announcement in a single course can trigger hundreds of thousands of backend jobs.
In spite of all these challenges, the system I built was still in use after the user base grew 40x.
Built the wishlist functionality and helped initiate some of the most lucrative email campaigns of Silicon Valley done at the time. We would send discount codes to users who added a course to their wishlist and didn't buy it in a month.
We were making around $1 per email on average.