Artem Shcherbina

I build and support document intelligence and workflow automation systems for European small and mid-sized businesses.
DocuCapture — a document capture, verification and data extraction platform for a legal-services business in Germany and Denmark, in daily production use.
The platform combines guided photo/PDF capture, automatic quality checks, OCR with user confirmation and staff review. It supp

• Developed document AI for Samsung Gauss; contributed to teams that placed first in the ICDAR 2025 literary epoch classification competition and second in the ICDAR 2024 multi-font OCR competition.
• Trained and deployed LLMs on multi-GPU clusters; optimized inference using vLLM, llama.cpp and quantization; built internal RAG assistants for 100+ employees.
• Developed gesture recognition features

• Conducted statistical analysis of government forest inventory data to inform policy decisions.
• Developed advanced prediction models for tree height and timber volume, enhancing accuracy in resource management.
• Created a comprehensive data analysis project using R, streamlining data generation processes.

• Applied statistical modeling and analysis to large-scale stratified government datasets
• Developed robust estimators for total enterprise turnover, employee counts, and income across multiple subgroups
• Constructed confidence intervals and uncertainty quantification for complex survey estimates
• Handled missing data, errors, and data cleaning to ensure high-quality inputs for predictive model