Sergei Ivanov

Also known as Sergey Ivanov

Sergei Ivanov is a machine learning researcher whose work connects recommendation systems with graph learning and language-model evaluation. He coauthored a SIGIR paper on viral-content recommendation in 2017, published anonymous walk embeddings in 2018, and completed a Skoltech PhD in Computational and Data Science and Engineering in 2019. At Criteo, his 2021 work paired gradient-boosted trees with graph neural networks. In 2022, he proposed compatible label propagation for heterophilous graphs and joined a KDD tutorial team on reward-optimizing recommendation. He coauthored SWE-InfraBench for evaluating language models on AWS CDK code changes, and as of October 2026 he is a Principal, Applied Scientist at Amazon.

Now at: Amazon
Works: 2
Citations: 0
h-index: 0
Badges: Bridge

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

2017

Content recommendation for viral social influence

At SIGIR in 2017, he and Konstantinos Theocharidis, Manolis Terrovitis, and Panagiotis Karras publish “Content Recommendation for Viral Social Influence.” The paper chooses content attributes for a campaign that starts with a fixed seed set and seeks to maximize network spread.
2018

Anonymous walk embeddings

In 2018, he and Evgeny Burnaev publish anonymous walk embeddings for unsupervised representations of entire graphs. Their SVM features increase graph-classification accuracy over convolutional neural networks and traditional graph kernels.
2019

A Skoltech PhD

At Skoltech, he earns a PhD in Computational and Data Science and Engineering in 2019 under the supervision of Associate Professor Evgeny Burnaev.
2021

Boost then Convolve

At Criteo AI Lab, he and Liudmila Prokhorenkova publish “Boost then Convolve” at ICLR in 2021. Their architecture jointly trains gradient-boosted decision trees on heterogeneous features with graph neural networks for graph structure.
2022

Compatible label propagation

In 2022, he, Zhiqiang Zhong, and Jun Pang propose compatible label propagation for node classification on heterophilous graphs. Their method combines a base predictor with label propagation and has a closed-form solution with convergence guarantees.
2022

A KDD recommendation tutorial

In 2022, he joins the KDD hands-on tutorial team for “Reward Optimising Recommendation using Deep Learning and Fast Maximum Inner Product Search.”
2025–2026

SWE-InfraBench for AWS CDK

He and 12 coauthors publish SWE-InfraBench at the NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle. Its 2026 arXiv preprint evaluates language models on realistic code changes in AWS CDK repositories.
As of October 2026

Principal applied scientist

As of October 2026, he is a Principal, Applied Scientist at Amazon.