At Victory Technology, I developed local NLP pipelines to analyze complex legal texts in Russian and Kazakh, extracting key provisions, arguments, and semantic relationships into structured JSON. I also built document-processing and quality-assurance workflows with normalization, rule-based extraction, regression tests, and manual-review diagnostics.
I developed computer vision components for object detection and visual inspection, and worked on model inference and integration into application workflows. In my projects, I developed a PatchCore and PaDiM visual anomaly-detection system and built a PyTorch Geometric approach to coordinating distributed PID controllers.
