Aidana B

Computer Vision Engineer

QazCode

IT holding of Beeline.

Works on biometrics, document OCR, identity fraud prevention, and wildfire detection, including an OCR model that improved 10% on internal benchmarks.

• Worked in a biometrics team on liveness detection and anti-spoofing systems to prevent presentation attacks in identity verification pipelines. • Contributed to document intelligence projects, including OCR-based document recognition and anti-fraud detection for identity documents. • Evaluated VLMs for document intelligence. Qwen 72B reached 95% accuracy, while other models showed weak reasoning in Kazakh. • Retrained the OCR model, improving accuracy by 10% on internal benchmarks, and deployed it in C++. • Deployed one-to-one face-to-ID photo verification for fraud prevention.

Research Assistant

Computer Science Department, Nazarbayev University

Nazarbayev University is a research university in Astana, Kazakhstan, with undergraduate and graduate programs.

Researches Kazakh language AI, including multilingual retrieval, cultural image generation, CLIP models, speech systems, and document-grounded chatbots.

• Designed and evaluated a large-scale retrieval system using FAISS and multilingual embeddings, measuring Recall@k and ranking performance across multilingual datasets. • Trained Stable Diffusion v1.5 with DreamBooth to produce Kazakh cultural style LoRA weights, curated a domain dataset, and delivered an internal demo. • Conducted a study of CLIP-style models for Kazakh, comparing an mBERT text encoder with ViT, EfficientNet, Tiny Swin, and ResNet-50 vision backbones. • Implemented a text-to-video retrieval pipeline using a CLIP model.

Computer Vision Intern

KarTel

Telecommunication company.

Built an OCR platform that converted documents into four formats and developed AI-generated avatars for internal business use.

• Collaborated with cross-functional teams to analyze data on scammers. • Implemented an OCR platform for colleagues to convert documents into four postprocessed formats. • Developed AI-generated avatars for internal business use.

Research Assistant

Institute of Smart Systems and Artificial Intelligence, KazLLM team

The Institute of Smart Systems and Artificial Intelligence is a research institute at Nazarbayev University focused on AI.

Improved Kazakh OCR accuracy to 93.1% with a custom dataset and reached 99.2% with OCR model optimization.

• Improved Kazakh language recognition accuracy from 46.77% to 93.1% by fine-tuning Tesseract OCR on a custom dataset of 10,000 scanned Kazakh text lines. • Researched and optimized hyperparameters in PaddleOCR, EasyOCR, and SuryaOCR, reaching a peak accuracy of 99.2% for Kazakh language recognition. • Collected a dataset of over 500,000 entries for retrieval-augmented generation using Selenium web scraping. • Recommended deep learning reranking models with a 97% hit rate, including Jina AI, ColBERT, Cross Encoders, and BAAI, to improve performance.

Data Science Assistant

Outpeer.kz

Programming school.

Worked as a Data Science Assistant at a programming school. The resume gives no further role details.

B.Science, Computer Science

Nazarbayev University

M.Science, Data Science

Nazarbayev University