Lay the foundation of ML development in the company: experiment tracking, reproducibility, model release, and quality monitoring;
Exploratory analysis on ClickHouse data (raw and curated layers) — trends, anomalies, and a clear picture of what powers the Intelligence product;
Design, train, and iterate models for Intelligence scenarios inside user cabinets: analytics and predictive product metrics (scoring, churn, forecasting, anomaly detection), personalization, and guided action scenarios;
LLM interface on top of analytical data: prompt design, evaluation, answer validation against ClickHouse and curated corporate data, with strict access-control compliance;
Feature engineering primarily in ClickHouse via SQL (aggregates and materialized views), avoiding unnecessary large-scale exports out of the cluster;
Ship minimal viable models; validate product impact versus heuristics and strong baselines;
Offline evaluation and online experiments (A/B or equivalent) on real traffic or users before full-scale rollout;
Partner with data engineers when new sources land or schemas evolve;
Align with BU product teams on hypotheses and priorities.
What we expect:
Strong Python for ML and analytics;
Advanced SQL for analytical extracts and feature engineering in ClickHouse (aggregations, window functions, performance-aware query design);
Hands-on experience with Pandas, NumPy, and scikit-learn — training, validation, interpretation, and product-aligned metrics;
Solid probability and mathematical statistics; hypothesis testing and A/B or quasi-experiments;
Understanding of common ML algorithms and their limitations;
Understanding of columnar analytical databases;
End-to-end experience shipping ML models to production (from data to serving), including environments without a pre-existing ML-Ops stack;
MLflow, DVC, or equivalent for reproducible experiments and clean handoff to engineering;
Understanding of version control systems (Git) and experience using Git in team-based development;
Hands-on experience with AI and LLM models via APIs — prompting, model comparison, and output-quality checks for Intelligence-style scenarios grounded in ClickHouse and curated corporate data.
Nice to have: Experience with RAG, embeddings, or fine-tuning.
We offer:
Work in the international company;
Hybrid working format (office/home office);
Corporate education — courses and trainings;
Voluntary health insurance after probation period;
