Dmitry Kan

May 23, 2023

14 stories

1 save

Approximate Nearest Neighbour Search: algorithms, products, tradeoffs, research and development

Presentation (together with Max Irwin) at NLP Meetup by Deepset, January 2022
Beyond KNN/ANN algorithms, that you can try in your ML / Data science lab, how about scaling this to a production database level?
12% improvement over FAISS on 10M vector scale -- was achieved as part of Billion-Scale competition with my team Sisu
Berlin Buzzwords conference is my favorite in Europe for many years (since 2011?) This year I've had the honor to co-present -- in the form of AMA session about vector search. Dive into 3 most interesting questions from the audience and my answers
This blog goes much deeper -- to the similarity search algorithm on Apache Lucene level. It upgrades the code from 6.6 to 8.0
Apache Lucene also powers Elasticsearch -- and we can finally also into a KNN/ANN space, i.e. we don't have to traverse the entire set of vectors to find the K similar ones
Dmitry Kan

Dmitry Kan

Founder and host of Vector Podcast, tech team lead, software engineer, manager, but also: cat lover and cyclist. Host: https://www.youtube.com/c/VectorPodcast