TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Show HN: Tamber – put real time, high-accuracy recommendations in your app

9 pointsby alexirobbinsover 7 years ago

2 comments

alexirobbinsover 7 years ago
Hi all! I&#x27;m Alexi, founder of Tamber. We built Tamber to help developers put fast, effective recommendations into their apps.<p>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 <i>is</i> similar to Bob Dylan, but that doesn&#x27;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.<p>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.<p>It works just like an analytics service, except that every event you track triggers a system-wide update to the model. And it&#x27;s really fast, returning fresh suggestions in 20-120ms. So as a user navigates around your app (even if they aren&#x27;t signed in!) your app can always display the optimal set of next things they should see next.<p>Here is a simple demo app for book recommendations we made using Goodreads data pulled from Kaggle: <a href="https:&#x2F;&#x2F;tamber.com&#x2F;demo&#x2F;goodbooks" rel="nofollow">https:&#x2F;&#x2F;tamber.com&#x2F;demo&#x2F;goodbooks</a><p>I&#x27;ll open source the app code once I clean it up a bit.<p>Looking forward to hearing your thoughts and feedback!
LoremTechover 7 years ago
Our company has been using this for a few months now and it’s seriously awesome - has saved our developers about 100 hours of work...paid for itself 10X over and seeing a bump in user time in product