Hi all! I'm Alexi, founder of Tamber. We built Tamber to help developers put fast, effective recommendations into their apps.
After trying a few open source libraries for a music app I was building, I found that they were surprisingly tedious to implement and tended to overfit for popular items – Neil Young is similar to Bob Dylan, but that doesn't help you discover new music. I knew there had to be a better approach that would solve this popularity bias problem, and make recommendations less painful to implement.
Tamber overcomes popularity bias by learning not only the relationships between items, but also how trends in taste evolve over time and using that information to boost less-well-known items in recommendations.
It works just like an analytics service, except that every event you track triggers a system-wide update to the model. And it's really fast, returning fresh suggestions in 20-120ms. So as a user navigates around your app (even if they aren't signed in!) your app can always display the optimal set of next things they should see next.
Here is a simple demo app for book recommendations we made using Goodreads data pulled from Kaggle: https://tamber.com/demo/goodbooks
I'll open source the app code once I clean it up a bit.
Looking forward to hearing your thoughts and feedback!
