High School

Master's degree

Doctor of Philosophy (Ph.D.)

Master of Science - MS

TA for Machine Learning course by Prof. K.V. Vorontsov

Teaching Assistant

Started the recommender systems direction at Yandex — designed core approach and built the first framework
Led development for recommendations in Yandex Music, Market, and Zen
Launched personalized radio and feed for Music, boosting engagement
Built “Users also bought” for Market, increasing conversions
Architected and shipped the first version of Zen — later became Yandex’s homepage feed

Technical Lead Manager

Optimized ranking objectives for Microsoft Start Feed, improving engagement and user satisfaction
Architected, improved, and shipped core ranking components using tree-based and neural models

Principal Applied Scientist

Built DJ — Yandex’s core recommendation platform powering 30+ products; acted as architect and engineering manager
Invented algorithms for generative ranking, transformer-based user modeling, retrieval quota optimization, and RL-based ranking
Drove multi-fold metric gains in music, video, product, and image recommendations
Owned end-to-end recommendations for Yandex Smart Devices
Standardized ML hiring by launching a unified interview process

Designed quantization methods for bin-counting models to reduce memory usage

Research Intern

Developed forecastability framework and automated model selection pipelines for time series

Built ALS-based recommendation library still in production across multiple products
Developed media sentiment analysis pipeline for brand monitoring
Created forecasting system using search index data to predict box office and startup success

Staff Software Engineer

Founding engineer on YT, Yandex’s core system for distributed data storage and batch
processing
Conducted research on recommendation models and built a film recommender prototype

Software Engineer

Created PoetryLover, a poetry quotation search engine launched publicly in 2009

Software Engineer Intern

Tech Lead for “For You” recommendations — X’s main engagement surface
Improved user retention and time spent via ranking and retrieval model upgrades
Shipped new retrieval models for “For You” and video
Invented unbiased in-batch retrieval evaluation, spaced repetition-based training
Developed joint retrieval + light ranking training
Enhanced transformer objectives and boosted diversity and reranking quality with listwise signals

Senior Staff ML Engineer

Member of Technical Staff

Senior Staff ML Engineer at X/xAI

substack

Know about Michael Roizner's work experience, education, bootcamps, certifications, and much more. Michael Roizner