Lead Data Scientist (Training of analysts + ML Delivery) in ATB-market

We invite you to join our team Lead Data Scientist

Role:

• Build analytics processes so that data is stable, reproducible and controllable; 
• start the training of analysts and ensure the delivery of ML solutions from problem setting to production.

Responsibilities:

• Data engineering & pipelines: setting up Dagster + dbt; data tests, alerts, leakage control.
• Analytical showcases: design of consistent marts (customer/check/product/store/promo/channel) with correct grain and historicity.
• ML for tabular data: building and validating models (LightGBM/XGBoost/CatBoost), regularization, CV, class imbalance work, interpretation (SHAP).
• Assessment of model quality: ROC-AUC/PR-AUC, F1, calibration and others.; preparation of metrics and reports for business.
• Full ML/DS cycle: task setting - preparation of datasets - modeling - interpretation - production (batch/API), Docker.
• Training/mentoring: system upskill training for analysts (Excel level and above), regular classes and review of tasks.
• Standards commands: Git, code review, notebook/report templates, documentation; implementation of the "Data Platform Playbook".
• Data mining: finding patterns and hypotheses on real data, working together with business.
• Additionally - Architecture and data platform: participation in the deployment of MinIO + Apache Iceberg + Catalog + Trino; ensuring data quality and manageability.

Requirements (technical):

1. Python + SQL (strong): pandas/numpy, scikit-learn; CTE, window functions, query optimization.

2.   Mathematical base (practical):

• probability and statistics: distributions, expectation/variance, confidence intervals, p-value;
• hypothesis testing, A/B tests, statistical power;
• linear algebra: matrices/vectors, basic understanding of gradients.
  
• 
  3.  ML for tabular data: LightGBM/XGBoost/CatBoost, regularization, bias-variance, cross-validation, leakage control.
  4. Evaluation of models: ROC-AUC/PR-AUC, F1, calibration; work with imbalance; interpretation (SHAP).
  5. End-to-end DS: from problem setting to production (batch/API), Docker.
  6.Training/mentoring: work with Excel level analysts; system classes + review.
  7. Upskill program: ability to design a plan for 3-6 months (practice/homework/skills matrix).
  8.Team standards: Git, code review, templates, documentation.

Will be a plus: experience with Lakehouse, Trino performance tuning, production-ML solutions in Retail/FMCG, CI/CD experience for DS.

Tasksfor the pilot (first 6 months):

1. Join the “data factory” deployment project (MinIO + Iceberg + Catalog + Trino) — ensure stability, reproducibility, control.
2. Build basic data windows for customer analytics (customer/check/product/shop/promo/channel) with agreed grain and historicity.
3. Set up automatic pipelines (Dagster + dbt), data tests and alerts.
4. Build processes of data processing and analysis, data mining.

Internal training (required):
5. Conduct a SQL Bootcamp for a pilot group (3-4 people): SELECT/JOIN/GROUP BY, window functions, grain logic, rules "how not to break metrics".
6. Create a “Data Platform Playbook”: how to connect, where which tables, what the “source of truth” is, how to request new fields/tables (application process).
7. Run office hours 2 times/week: analysis of real tasks of analysts on real data.

The company offers:

• remote or hybrid format work;
• employment on the terms of a gig contract or in the state (reservation is possible);
• paid annual leave of 24 calendar days, paid sick leave;
• regular payment of wages without delays and in stipulated amounts, regular salary review;
• opportunity for professional and career growth;
• training courses.

Contact person: Kateryna, tel. style="font-weight: 400">0984567857 (t.me/KaterynaB_HR)