Venu Satuluri

I am currently exploring new opportunities. I am very interested in building opinionated products that focus on impact for the end user.

From 2012 to 2022, I worked at Twitter creating new products and systems in the space of recommendations and machine learning. Six months after I joined, me and a few team-mates started a new product called MagicRecs for real-time recommendations. I was the lead engineer for MagicRecs and helped it become a key growth driver for Twitter. In 2017, I created a new technology called SimClusters, which enabled a quantum leap in Twitter’s ability to model user interests and spin up new recommendations products effortlessly. At Twitter, I was fortunate to work with wonderful colleagues who trusted my ability to rise to new challenges. I held the role of a Principal Software Engineer at the time I left.

Previously, I did my Ph.D. at Ohio State University. My favorite papers from my Ph.D. is the work on BayesLSH and the work on Directed graph symmetrizations, both of which influenced the SimClusters work I did later at Twitter.

Some recent analyses:

Comparison of different transformer implementations for small applications From April '23
A surprising observation on OpenAI's text embeddings From May '23
Comparison of different ways to embed nodes in a graph From Dec '22
Evolution of female representation in Bollywood, 1947-2022 From May '23

Selected Github Repos:

Various analyses (mostly from 2023), including the ones above
SBF (Sparse Binary Factorization), a package for community discovery from billion-node graphs, used in SimClusters.

Selected papers (full list at my Google Scholar page):

Simclusters: Community-based representations for heterogeneous recommendations at Twitter. KDD 2020
MagicRecs: Real-time twitter recommendation: Online motif detection in large dynamic graphs. PVLDB 2014
Factorbird: a parameter server approach to distributed matrix factorization. DML workship, NIPS '14

Venu Satuluri

Timeline