Andrei Kalenik Email & Phone Number

Krakowski, Małopolskie, Poland

Minsk, Belarus

PyTorch course ZeroToMastery• Created sophisticated models for food, clothing, and numerical image classifications, yielding accuracies ranging between 97-98.5%.• Increased overall accuracy by 8-9% through effective utilization of data augmentation techniques.• Acquired comprehensive knowledge of state-of-the-art machine learning algorithms.Machine Failure Prediction• Secured a position within the top 3% in a Kaggle competition• Developed a predictive model for machine failure based on parameters such as current CPU and air temperature, usage time, etc., achieving an AUC-ROC metric of 0.98.• Enhanced model performance by data preprocessing. To address the unbalanced dataset, I implemented oversampling, culminating in a total AUC-ROC metric improvement of +0.18.Reinforcement Learning Snake Game• Leveraged reinforcement learning principles to develop an autonomous Python-based snake game, achieving a minimum score of 20 points post-training.• Employed policy optimization techniques to enhance the game’s performance, resulting in a remarkable mean score increase of 10 points.Protein Dataset Analysis• Conducted a comprehensive analysis of a protein dataset to identify the most influential features. Leveraging this analysis, I successfully trained a model to predict protein types, achieving an impressive AUC-ROC score of 0.96.• Employed sklearn library to optimize the model’s hyperparameters. This resulted in a growth of AUC ROC score, with an improvement of +0.12. Additionally, the training time was significantly reduced by 56%.Python Game Development Expertise: Neon Square and Pixel Fight• Utilized Python’s OOP capabilities to program engaging games - Neon Square and Pixel Fight, available on Play Market. These games have amassed 1500+ downloads, full self-deployment from start to finish.• Implemented efficient code optimization techniques to ensure compatibility across various devices, boosting the mean frames per second by 40%.

Minsk, Belarus

• Developed two distinct algorithms for opening trades. Made the first algorithm to utilize the price on two differentcharts, resulting in a 80% total profit. Implemented the second algorithm to focus on analyzing price changes overtime and has successfully achieved a 70% profit.• Analyzed price changes and market trends over the last 2 years, identified a pattern that helped create a modelthat when tested over the last 2 years brought 120% profit from initial deposit.• Mastered the art of customer engagement, evidenced by the delivery of more than 70 successful orders.