Nikodim Svetlichnyi

Designed and deployed a production ensemble (XGBoost/CatBoost/LightGBM) achieving F1=0.84 and ROC-AUC=0.89 on 118K imbalanced samples, reducing false positives by 38% while serving 100K+ daily predictions with 99.7% uptime. Optimized feature pipeline to cut inference latency 2.8×, reduced feature space from 43 to 7, and delivered a 15% operational cost reduction for real-time predictions.

Developed LSTM forecasting (2 layers) improving prediction accuracy 18% vs baseline and an autoencoder for anomaly detection achieving 83% recall with 5.1% FPR. Implemented hybrid clustering and regression pipelines to enable segmentation and reduce feature space while maintaining model performance.

Built a recommendation engine (collaborative filtering, SVD) achieving Precision@10=0.31 and increased engagement 26% in pre-release testing; combined logistic regression and XGBoost for classification achieving ROC-AUC=0.76 and improved conversion prediction by 28%.

Developed scoring models using weighted linear regression and random forest achieving R2=0.64 and built automated ETL pipelines processing 50K+ daily records with validation and outlier detection. Implemented Monte Carlo simulations (10K iterations) to calculate VaR and CVaR for portfolio risk assessment.

Grade: 7.83 ! ACCELERATED 1-YEAR TRACK, FULL 2-YEAR CURRICULUM !

Activities and societies: Master program on Artificial Intelligence, Robotics & Autonomous Systems; Master Thesis: Robustness in Reinforcement Learning under Task-Uncertainty.

Completed an accelerated M.Sc. in Intelligent Interactive Systems covering AI, robotics and autonomous systems with a thesis on robustness in reinforcement learning under task uncertainty.

Grade: 7.62

Activities and societies: Sailing Club Hermes; Bachelor thesis: Product Category Analysis and Optimization for the Lenta Online Store.

Completed a Bachelor of Economics with strong quantitative training in mathematics, statistics and econometrics and applied data analysis in Python; bachelor thesis on product category analysis and optimization.

Activities and societies: Practical training in AI fundamentals, hands-on Python exercises and practical case analyses.

Completed a short course on Deep Learning & Neural Networks covering practical machine learning and deep learning fundamentals with Python.