Junior Data Scientist; Onsite position

About the role

The project is dedicated to building advanced data science solutions to improve the integrity and efficiency of our operations across various domains. This includes developing systems for anomaly detection, fraud prevention, predictive maintenance, and customer lifecycle management.

Initially, you will focus on preprocessing, cleaning, and validating data across multiple sources to support these initiatives. Over time, you will take ownership of model development, monitoring, and improvement across a broad range of applications:

• Anomaly Detection:
  Identify unusual patterns and behaviors using supervised, unsupervised (e.g., Isolation Forest), and sequence models (e.g., HMM) to mitigate various risks.
• Behavioral Analysis:
  Detect deviations from expected behavior and identify unauthorized actions through transaction scoring, reconciliation, and peer baselining.
• Performance Monitoring:
  Spot abnormal consumption and frequency patterns across different entities using anomaly detection and change-point detection.
• Asset Management Decisions:
  Inform repair or replacement decisions for assets using survival/RUL models and cost-sensitive classifiers.
• Customer Engagement:
  Predict changes in customer engagement and trigger appropriate interventions (e.g., repossession, incentives) using churn classifiers, causal uplift models, and customer segmentation.
• Customer Lifetime Value:
  Guide thresholds for engagement and retention strategies with survival analysis and gradient models based on customer behavior.

• Data Management & Preprocessing
  • Collect, clean, and preprocess data from multiple internal and external sources.
  • Ensure data quality, consistency, and integrity for model input.
  • Handle missing values, anomalies, and inconsistencies in large-scale operational datasets.
• Model Development & Ownership
  • Support the design, training, and validation of predictive models.
  • Run experiments with different algorithms (logistic regression, decision trees, gradient boosting, etc.) and compare performance.
  • Gradually take ownership of existing models, monitoring their performance and leading refinements.
• Collaboration & Reporting
  • Work closely with the data science lead to implement improvements suggested by exploratory research.
  • Summarize findings and model performance for both technical and non-technical stakeholders.

• Strong foundation in data analysis and statistics (e.g., distributions, correlations, hypothesis testing).
• Proficiency in Python (pandas, Num Py, scikit-learn) and SQL for data manipulation.
• Experience with data visualization libraries (Matplotlib, Seaborn, Plotly, or similar).
• Understanding of supervised and unsupervised learning methods (logistic regression, decision trees, clustering, anomaly detection).
• Ability to handle messy real-world data and apply rigorous quality checks.
• Ability to communicate effectively in Russian

• Familiarity with gradient boosting libraries (e.g., XGBoost, Cat Boost) and model evaluation techniques (precision, recall, F1, AUC).
• Knowledge of feature importance and explainability methods (e.g., SHAP values).

• Analytical curiosity with a problem-solving mindset.
• Strong communication skills to present findings clearly.
• Ability to work independently and take responsibility for deliverables.