Shaped

How can large language models (LLMs) be used to help your users find what they're looking for? In this post we dive into a survey paper that discusses this topic. It looks at the benefits of using Large Language Models for content/product recommendations. And highlights many of the challenges of integrating these models in a production setting (inference latency, bias, long-text input, etc...). At Shaped we've been exploring the benefits of these models also. We're excited about seeing a unified evaluation benchmark for this line of work, and generalized recommendation foundation models that work across domains. #ai, #ml, #recsys

Got customer interaction data? You can build a recommendation system. To find out more, check out the 'What is Shaped' page in our docs here: https://lnkd.in/gT85puYW

We recently talked to James A. Rolfsen, previously Director of Data Science at Rappi, about the difficulties of building recommendation systems at scale, and how companies can leverage their existing data using off the shelf solutions to create powerful recommendations and search. His take - even messy, unstructured events data can be incredibly valuable. By leveraging customer interaction data and applying representation learning to create embeddings based on cumulative interactions, companies can get high-signal representations of user preferences. This data can then be used to optimize recommendations and search. Many companies don't realize the potential of the data they have. We'll be sharing more insights from James, as well as other industry experts over the coming weeks 👀

"You need a team of 10, 20, 30 people to build recommender systems at scale. If you've got an off the shelf solution that can do that, if you can take all that out of the equation, it's unreal." We recently chatted to John Hannebery about the difficulties of building recommender systems at scale, the timelines involved, and the headcount required. With Shaped, one engineer can implement a state-of-the-art search and recommendation system in as little as a day. For more information on Shaped’s ease of implementation and configurability check out our quickstart here: https://lnkd.in/g24324Tj

Check out our latest case study on how podcast app Overlap (YC S24) leveraged Shaped to solve the cold start problem and deliver personalized podcast recommendations from day one: https://lnkd.in/gk6-vQQy Overlap’s primary goal was to improve the discoverability of podcasts content with feed personalization. For a new startup, creating a curated feed at a user level is made difficult due to the fact they face the cold start problem for 100% of their users. By partnering with Shaped, Overlap was able to provide more personalized and diverse content recommendations, promoting visibility for new content and enhancing user engagement from the outset.

👏🏼

Shaped ranked #5 on Product Hunt! Thank you to everyone that took the time to support us 🤝

Exciting times!

Shaped is currently ranked 5th on Product Hunt, and is only 15 upvotes away from 4th. If you'd like to support us, check out our Product Hunt page here: https://lnkd.in/gqC89vSz

We are LIVE on Product Hunt 🎉 Check out our Product Hunt page and give us some support here: https://lnkd.in/gqC89vSz