Junior CV Engineer
PBSOFT LLC
Jul 2024 — PresentBuilt document and identity computer-vision workflows involving detection, OCR, APIs, and browser-based ONNX inference; the document-processing integration reduced manual moderation.
• Collected and annotated a dataset, and trained YOLO to detect key elements (seals, coats of arms, inscriptions) for a draft document classification system for the Ministry of Defense of the Republic of Kazakhstan. • Developed a FastAPI REST API to process multipage PDF, JPEG, and PNG files. The solution was integrated into the system and reduced manual document moderation. • Fine-tuned Tesseract OCR on a custom dataset with Kazakh and Russian Cyrillic. • Developed a synthetic data generator for IIN, ID, dates, and full names, varying font, lighting, and noise.