The Center for Practical AI is a division of Sber that works on complex AI projects and solves non-trivial tasks for the bank and the Sber ecosystem.
Our team builds and fine-tunes an LLM lineup for banking scenarios and deploys models in production within "strict" environments: local devices / closed networks / strict latency SLAs. The work involves a full cycle: data → fine-tuning → quality assessment → inference optimization → load testing → deployment.
The first stage of selection for this vacancy is a conversation with an AI recruiter. After applying, you will receive an invitation to your email for a preliminary interview with GigaRecruiter in Telegram. The conversation will take approximately 10 minutes. Its purpose is to clarify missing details and expedite the review of your candidacy. GigaRecruiter is just starting out, so we ask for your understanding. Your experience and participation will help make it convenient and useful!
You will:
- train LLMs on GPU clusters
- collect, clean, and prepare datasets for pre-training and fine-tuning
- analyze experiment results and improve model quality based on metrics
- optimize inference, conduct load testing, and deploy trained models.
We expect you to have:
- confirmed practical experience in training LLMs (pre-training, fine-tuning, alignment) in research or production tasks
- a deep understanding of theoretical and practical aspects of LLMs: architecture, tokenization, pipeline construction, batching, mixed precision, distributed training, debugging failed training iterations
- 3+ years of Python development experience, writing clean and maintainable code within modern ML projects
- attention to detail and good communication skills.
It will be a plus if you:
- have developed AI agents (tool-using, planning, multi-step workflows), are familiar with agent frameworks and architectures
- know how to use experiment logging tools (Weights & Biases, MLflow, Langfuse, etc.)
- are proficient in inference optimization and model hosting in production
- have designed and maintained eval pipelines for LLMs with metrics, dashboards, experiments, and regression checks.
We offer:
- a comfortable modern office in Moscow, near Kutuzovskaya metro station
- the opportunity to choose a convenient schedule – office/hybrid (with at least 2 days a week in the office)
- annual salary review and annual bonus
- corporate gym and relaxation areas
- over 400 educational programs from SberUniversity for professional and career development
- voluntary medical insurance, preferential insurance for family, and a corporate pension program
- a flexible discount on mortgage loans equal to 1/3 of the Central Bank's key rate
- a free SberPrime+ subscription, discounts on products from partner companies.
