Project: AI AI-Engineer

Project: JetBrains Academy - a project-based learning platform where users learn to program by creating real-world applicationsKey achievements:• Designed and implemented a pipeline for AI Generated Content (AIGC) utilizing the GPT model, successfully generating unique code and Faded Parsons problems to enhance programming learning experiences.• Developed an adaptive learning system based on Large Language Models (LLM), increasing the user retention rate by 20%.• Provided mentorship and guidance to a team of machine learning interns, fostering their professionaldevelopment.

Project: AI-powered Home GymKey achievements:• Created Computer Vision models for pose estimation and object detection using TensorFlow and TFLite achieving AP metric 0.73 on the gym exercises dataset.• Integrated Computer Vision models into iOS application and C++ application.• Implemented function resize with padding in the iOS application, reduced operation time by 3 times.• Optimized h264 video encoding process by transferring to GPU. Reduced memory usage twice.

Project: Recommendation system for BurgerKing RussiaDesigned and deployed recommendation system for Burger King restaurants using NLP models andcollaborative filtering. Improved recommendation-based sales in more than 200 stores across Russia by 90%.

• Developed and deployed NLP-based payment classification system for the largest bank system in Russia.• Optimized determining the price of real estate in Moscow by combining several classical ML approaches. Improved the developer’s profit margins by20%.

Tomsk branch of Heriot Watt UniversityKey achievements:• Used transfer learning with generative models (GAN) for petroleum parameters generations. Maps generated by GAN correspond to the properties of the dataset by 97%.• Enhanced the algorithm for finding a suitable location for drilling a well based on grid data by using multiprocessing. Reduced calculation time from 40 to 2 minutes.• Developed .NET application for estimating objective function coefficients based on grid data. Minimizedhuman involvement in the process, reduced the time from 20 minutes to 5 seconds.• Produced PyQt application for estimating the indicators of porosity and permeability at different depthsbased on well justfications.