Data Scientist in the Deep Personalization Team

About the Team

We are responsible for personalizing Avito's main page — selecting listings for tens of millions of users so that each person sees the most relevant content. We develop two-tower transformer models, work with state-of-the-art approaches in recommendation systems, and handle the full development cycle: from researching architectures to production inference with ANN search and embedding caching.

An example of our work: https://www.youtube.com/watch?v=Fn-mTFuRpHU — a talk at DataFest about two-tower models for personalization.

Examples of future tasks:

You will:

• design and maintain Spark ETL pipelines for processing billions of events;

• build data marts and manage their quality;

• optimize distributed training dataset preparation;

• develop the architecture of a two-tower transformer model: improve user and item encoders, experiment with image processing and loss functions;

• research and adapt modern approaches: sequence modeling, multimodal encoders, advanced retrieval architectures;

• scale training to multi-GPU/multi-node, optimize throughput and convergence;

• integrate models into production: export to ONNX, configure ANN indices, work with Redis;

• conduct A/B tests, analyze impact on metrics (CTR, conversions, retention);

• iterate based on experiment results.

We expect you to:

• have a solid understanding of deep learning: have trained transformers, seq2seq, or two-tower models in production;

• be proficient in PyTorch and have experience working with large volumes of data;

• understand distributed training (DDP, FSDP) and know how to debug pipelines;

• have experience formulating hypotheses, designing experiments, and interpreting A/B tests.

It would be great if you:

• have experience in the field of recommendation systems or information retrieval;

• work with PySpark/SQL for big data processing;

• are familiar with modern research in the field of recommendations (Tiger, DSSM, contrastive learning).

Working with us means:

• the opportunity to influence business and product development;

• interesting and diverse tasks: analysts at Avito search for business growth points, study user behavior, develop frameworks, and set up dashboards;

• abundant high-quality data, powerful infrastructure and tools, any necessary hardware — everything is ready for productive work;

• a talented team, a great analytical culture, and a community of professionals;

• a transparent bonus system, a competitive salary — the exact amount will be discussed during the interview;

• a personal training budget that can be spent on books, courses, and conferences;

• care for your health: from day one you will have comprehensive health insurance including dentistry, with a therapist and massage therapist available at the office;

remote work and a wonderful office two minutes from Belorusskaya metro station: a panoramic view of the city center, spaces for focused work, and relaxation areas.