Andrei Kavalenka

Developed an intelligent agent as part of a multi-agent system for analyzing tabular data, generating over 5 types of visualizations and 10+ statistical metrics. Designed a retrieval-augmented generation (RAG) agent to extract and structure information from ~100 PDFs, enhancing search accuracy. Created forecasting models for business variables, mainly focusing on time series data, achieving up to

Developed an LLM-powered tool for automated job-candidate matching, enabling structured comparison across multiple criteria. Designed and implemented an LLM parser to extract key job requirements from vacancies and built a parallel LLM-based pipeline for candidate profile parsing. Implemented a weighted scoring system and ranking logic to shortlist top candidates, delivering structured summaries f

Developed an end-to-end internet TV ratings prediction project using Python and Bash, boosting business metrics up to 40% and incorporating MLOps features. Improved model predictions by 50% through monthly retraining, error correction, and refining preprocessing/postprocessing pipelines. Constructed data pipelines on AWS S3 and designed a Bash pipeline for manipulating Python projects on AWS EC2,

Applied various computer vision and custom methods to recognize italic and bold text with accuracies of 83% and 78% on customer's test dataset. Designed custom functionality to structure tables, paragraphs, and detect underlined/strike-through text in customer's PDF files, achieving 85-97% recognition accuracy. Linked C++ and Python using Ctypes and C++ libraries for parsing PDF structures, result

Studied Information Technology and Control, gaining foundational knowledge in various aspects of IT and system management. Focused on the principles and applications of technology and control systems.