Рабочий

Daniel Rabochyi

Data Scientist | Retail & E-commerce ML | CVM, Forecasting, Personalization

📞 [open contact info](look above in the "contact info" section)
✉️ [open contact info](look above in the "contact info" section)
📍 Kyiv, Ukraine

Profile

Data Scientist with hands-on experience in retail and e-commerce analytics, machine
learning, and automation of data/ML processes. Worked on solutions for customer churn
prediction, uplift modeling, promo demand forecasting, personalization, customer
segmentation, and basket analysis.

Experienced in full-cycle ML development: from data collection, feature engineering, and
model training to regular scoring, BI reporting, documentation, and delivering analytical
results to business teams.

Technical Stack

Programming & Data Analysis: Python, SQL, Pandas, NumPy, Scikit-learn
Machine Learning: LightGBM, CatBoost, XGBoost, Optuna
Big Data & Data Engineering: PySpark, Spark, Hadoop/HDFS, Hive, Iceberg, Trino
Databases: PostgreSQL, MS SQL Server
Automation & Analytics Engineering: dbt, Airflow
BI & Tools: Sisense, Git, Jupyter

Professional Experience

Data Scientist — Numbers

Worked on retail analytics and customer behavior modeling projects for Ukrainian and
international retailers. Focused on customer segmentation, store analytics, basket analysis,
churn prediction, and BI reporting for business teams.

Built customer segmentation models based on purchasing behavior, purchase
frequency, basket structure, promo sensitivity, category preferences, and customer
value.

Developed store and point-of-sale clustering approaches to identify similar trading
locations, compare performance patterns, and support business decision-making.

Conducted basket analysis and product similarity research using co-occurrence logic,
Yule’s Q similarity, and hierarchical clustering with Ward linkage.

Prepared analytical dashboards and visual reports in Sisense to monitor customer
behavior, store performance, product groups, promo activity, and key business KPIs.

Developed churn prediction models using transactional and behavioral features to
identify customers with a high probability of inactivity.

Translated analytical results into business recommendations for retention, targeting,
assortment analysis, and customer communication.

Data Scientist — EVA

Worked on promo demand forecasting and promo analytics for retail planning. The main goal
was to improve demand estimation for promotional campaigns and support procurement,
stock planning, and promo performance analysis.

Developed promo demand forecasting models at product, category, date, and promo
mechanics level.

Prepared historical sales, promo calendar, price, discount, product, category, and
seasonality features for model training.

Used machine learning models such as LightGBM and CatBoost for forecasting
promotional sales and estimating expected demand under different promo conditions.

Applied feature engineering for promo mechanics, discount depth, seasonality,
weekday effects, product hierarchy, historical sales dynamics, and lag-based demand
signals.

Used Optuna for hyperparameter tuning and model optimization.

Applied ABC segmentation and business-oriented error analysis to better evaluate
model quality for high-value and high-volume products.

Data Scientist — Fozzy Group

Worked on full-cycle ML and analytics automation projects for customer retention, uplift
modeling, recommendations, and personalized product ranking. Projects were implemented
with regular automated pipelines using dbt, Airflow, SageMaker, Spark, Trino, and Iceberg.

Developed Churn Prediction 2.0 based on snapshot logic, behavioral and transactional
features, promo activity, communication history, and customer lifecycle signals.

Trained and evaluated churn models using LightGBM, ROC-AUC, lift analysis, decile
analysis, and business-oriented validation.

Automated churn scoring pipelines with dbt models, scheduled Airflow DAGs, Spark
processing, Iceberg tables, logging, and regular scoring outputs for business teams.

Built uplift modeling solutions for promo targeting using Treatment/Control logic,
Two-Model Approach, XGBoost, Optuna, Qini coefficient, and decile-based business
interpretation.

Developed recommendation system pipelines for personalized product
recommendations based on customer-product interactions, purchase history, category
behavior, and product-level features.

Worked on personalized product ranking for e-commerce carousels by preparing user ×
product × carousel × date datasets and aggregating behavioral, transactional, and
contextual ranking features.

Trained learning-to-rank models using XGBoost Ranking and evaluated
recommendation quality with NDCG@5 and NDCG@10.

Built automated data and ML workflows with dbt, Airflow DAGs, SageMaker pipelines,
Trino queries, Spark jobs, and Iceberg-based feature tables.

Prepared production-like processes for regular feature calculation, model scoring,
result logging, and downstream usage in BI, CRM, and personalization workflows.

Education

Bachelor’s Degree in Computer Engineering

Taras Shevchenko National University of Kyiv
08/2021 – 06/2023

Focus areas: machine learning, model development, model training, hyperparameter tuning,
model evaluation, and performance analysis.

Master’s Degree in Computer Engineering

Taras Shevchenko National University of Kyiv
09/2024 – 05/2025

Thesis topic: Customer Churn Prediction in Retail

Worked on customer churn prediction models, training dataset preparation, feature
engineering, model evaluation, and analysis of customer retention strategies.

Languages

Ukrainian — Fluent
English — B2
Russian — Fluent