Todd Motto
@toddmotto · Founder, Voux.io
I’ve been alpha testing Enki for a little while, after helping out and contributing to the Angular material. Really useful sidekick of an app, looking forward to how this evolves further! One question I did have: how are you deciding what the three recommended options to learn today should be for me? Cheers guys keep up the great work! :)
8sharetweet・
Mathieu Dutour
@mathieudutour · Software Engineer @Enki
Thanks @toddmotto!
The recommendation is based on user preferences and previous interaction with the app.
We’re getting very fine-grain feedback from users because of the swipe interaction on each insight (one side for “learned something”, the other for “already knew”). This help us infer roughly the level of each user in each topic/subtopic, and we use that to predict what’s the most sensible insight to give them next. We’re also tracking the time spent on each insight, which is a good proxy for the level of interest and relevance of each insight for a specific users.
And finally, we learn about the users when they choose to do or not do a specific workout that we recommended. For instance - if a user was offered a “React” workout but chose not to do it, we’ll try next to offer him some “Angular” workout (or the other way around!).
This hopefully gives you an idea of the data we use in our Recommendation algorithm. As per the details of the algo itself: we’ll blog about it when it’s a bit mature and stable ;)
8sharetweet・