Nikita Melentev

Angestellt, Lead Data Scientist, Leroy Melrin

Moscow, Moscow, Russia, Russische Föderation

Über mich

Lead Data Scientist with 5+ years of experience. I have a focus on: • Leading ML teams, mentoring and interviewing DS specialists (4 teams, 100+ interviews and 7+ mentee) • Recommender Systems (Developed ML algorithms that increased turnover by 2.5% at Leroy Merlin) • Time Series Forecasting (Created 15+ new ML models at X5 Retail Group) • AB Testing and MLOps (10+ conducted online / offline AB-tests and 7+ fully finished ML projects)

Fähigkeiten und Kenntnisse

Data Science
Maschinelles Lernen
Python
Datenanalyse
Big Data
SQL
Statistik

Werdegang

Berufserfahrung von Nikita Melentev

  • Bis heute 2 Jahre und 6 Monate, seit Jan. 2022

    Lead Data Scientist

    Leroy Melrin

    Leading 10+ DS / DA specialists on 3 data projects: - Customer churn prediction model which improved retention by ~20% - Forecast of new items which increased items coverage by more than 100% - Computer vision models at the warehouse to scan QR codes and calculate cargo volume. Also: - Interviewed 100+ and hired 10+ DS / DA employees. - Mentored 7+ Junior / Middle DS / DA members of the team. - Managed 5+ GitHub repositories, checked work results, conducted code review of team members.

  • 1 Jahr und 3 Monate, Nov. 2020 - Jan. 2022

    Senior Data Scientist

    Leroy Merlin

    I worked on p2p recommender system for the website: - Created complementary goods generation and ranking algorithms. - Designed and conducted 3 online and 3 offline AB tests to check algorithmic improvements. - Grew complementary goods conversion to cart by 10% and turnover by 2.5% - Moved models to production with DE / DevOps colleagues using Airflow / DVC / Git / S3 / Greenplum

  • 3 Jahre und 5 Monate, Juli 2017 - Nov. 2020

    Data Scientist

    X5 Retail Group

    I worked in a sales forecasting team: - Replaced old ARIMA models with Prophet / Xgboost, added features like promotions, holidays, geolocation etc. This reduced forecasting error from ~25% to ~9% (MAPE). - Developed 15+ sales forecasting models for different planning horizons and product levels, including promo / holidays forecast, new item / store / region forecast, cannibalization effect analysis.

Sprachen

  • Englisch

    Fließend

21 Mio. XING Mitglieder, von A bis Z