Нікіта

NIKITA
PEREHNIAK
DATA ANALYST | DATA
SCIENTIST

CONTACT ABOUT ME
Data-driven engineer with 2+ years in Python/SQL and 1+
Valencia, Spain
year in ML. Delivered 5+ ML solutions boosting KPIs by
[відкрити контакти](див. вище в блоці «контактна інформація») up to 25%. Built ETL pipelines (2M+ records/day) and
[відкрити контакти](див. вище в блоці «контактна інформація») deployed 10+ REST APIs with Docker & CI/CD. Skilled in
cloud (AWS, GCP), ML, backend, and data visualization.
Kyivnik/github
Kyivnik/linkedin WORK EXPERIENCE
DATA SCIENCE INTERN (JAN 2025 - MAY 2025)
EDUCATION UnIP Ukraine
BACHELOR - SYSTEM ANALYSIS Developed LSTM-based forecasting model,
Kyiv National Economic University increasing weekly demand forecast accuracy by
2021-2025 12%.
Engineered SQL/Python pipelines to clean &
ERASMUS - MATH & C.S. aggregate 500K+ sales records/week.
VSE Prague Delivered interactive Tableau dashboards, reducing
2022-2023 manual report generation time by 70%.

SKILL DATA SCIENCE CONTRIBUTOR (OCT 2024)
NASA Space Apps Challenge
Machine Learning & AI:
Scikit‑learn, TensorFlow, Assisted in developing Exodoo, focusing on its AI
PyTorch, XGBoost, functionalities for exoplanet education.
hyperparameter tuning Managed database systems and executed
Data Engineering: Python ETL, machine learning processes for platform
Apache Airflow, BigQuery, enhancement (LangChain).
PostgreSQL, MySQL, data Parsed and integrated data from NASA and other
warehousing space organizations, maintaining data integrity.
Backend Development: Flask,
FastAPI, RESTful APIs, Docker, AI/ML DEVELOPER
GitHub Actions
Smart Grid Energy Anomaly Detection
Cloud Platforms: AWS (S3,
Lambda, EC2), GCP (BigQuery, Built a robust anomaly detection system using
Cloud Functions), Kubernetes Autoencoder, Isolation Forest, PCA, One-Class
basics SVM, and LOF for smart grid stability.
Data Analysis & Visualization: Delivered a real-time Streamlit app with instant
Pandas, NumPy, Matplotlib, MSE-based alerts and actionable visual insights for
Tableau, Google Data Studio operators.
Tools & Methodologies: Achieved >95% detection consistency across
Jupyter, Git, Jira, Scrum/Agile models, enabling up to 7x faster anomaly
response vs baseline methods.