Company:** Samokat (OOO Umnyy retail) Location: Moscow Salary: from 250,000 RUR Experience: 3 to 6 years

What you will be doing:

• Be responsible for ML components of products: classification, personalization, ranking, clustering — from task definition to deployment and monitoring;
• Develop the RAG system: improve knowledge base search quality, relevance of answers, control hallucinations, develop text annotation and indexing strategies;
• Build LLM quality evaluation from scratch: create datasets, implement automatic metrics, organize manual annotation, and track degradation after updates;
• Develop and test system prompts, measuring the effect of each change;
• Conduct product analytics: analyze funnels, cohorts, retention, A/B tests — and translate findings into concrete product decisions;
• Reduce the cost and latency of LLM calls through routing by task complexity, context compression, and caching;
• Work with multimodal models: perform image analysis and extract structured data from visual content.

What we expect from you:

• 4+ years of experience in Data Science / ML;
• Proficiency in the full ML cycle: feature engineering, training, validation, deployment, drift monitoring;
• Strong SQL skills and the ability to independently build product analytics without involving a separate analyst;
• Proficiency in Python at the level of production code (pandas, scikit-learn);
• Experience with embedding models (building and applying in search / RAG);
• Practical experience in evaluating LLM quality — created hands-on, not through ready-made wrappers;
• Ability to work with funnels, cohorts, retention, A/B tests and link results to product decisions.

Will be a plus:

• LLM fine-tuning (LoRA, QLoRA);
• Understanding of agent systems: ability to determine where the decision is made by the model and where by deterministic code;
• Experience with multimodal embeddings (CLIP, SigLIP and similar) for image search;
• Understanding of AI product unit economics: cost per query, optimization under load growth;
• Development of internal tools for model quality evaluation within the team.

What we offer:

• Five-day work week, eight-hour working day;
• Employment according to the Labor Code of the Russian Federation from day one and all social guarantees;
• White salary and Sber's salary project, thanks to which you will have access to preferential mortgage (or refinancing of an existing one) for those who have been with us for six months or longer, as well as a free Prime+ subscription, access to online cinemas, audio services, pharmacies, and more;
• Voluntary medical insurance with dental care after three months of work. The policy also includes online doctor consultations, travel insurance, and other services;
• Necessary equipment is provided for work;
• Conditions for professional development and career growth;
• Comfortable offices in cities of presence — with a quiet zone for work and meeting rooms for discussions and brainstorming sessions;
• Employee bonuses: corporate discounts from partners, contests and quizzes with prizes, gifts for children, and much more.