Matryoshka Representation Learning is a cutting-edge technology, positioned to transform Vector Embeddings and Search! The core idea is to learn multiple vectors of increasing lengths, which opens up all sorts of new opportunities for the mechanics of nearest neighbor search, the core technology behind Vector Databases!<p>I am SUPER excited to present our 89th Weaviate Podcast with Aditya Kusupati, the lead author of Matryoshka Representation Learning, Zach Nussbaum, a machine learning engineer at Nomic AI, one of the first companies to deliver Matryoshka embeddings in an API (paired with amazing efforts on open-sourcing the research and development), and Zain Hasan, a developer advocate at Weaviate who kicked off the podcast discussion with his amazing research on MRL and what it will mean for Weaviate!<p>I am so excited to present this podcast! It was such a fun conversation and is absolutely packed with information about these embeddings from what it takes to optimize them, interesting properties of variable-length embeddings, how Matryoshka embeddings fits in the Nomic AI ecosystem with products such as Nomic Atlas visualization, differentiable ANN indexes, and many more! I hope you find it useful!<p>YouTube: https://www.youtube.com/watch?v=-0m2dZJ6zos<p>Spotify: https://spotifyanchor-web.app.link/e/k3aqaTIylHb