Kazakh Handwritten Text Recognition Using Computer Vision and Neural Network
Jun 1, 2023 · International Conference on Electronics, Computer and Computation · Citations: 1
Machine Learning Engineer, Computer Vision and Applied AI
Computer Vision Engineer with nearly 4 years of experience. I have experience in Biometric Systems, Automatic Moderation and Medical Image Analysis.
Jun 1, 2023 · International Conference on Electronics, Computer and Computation · Citations: 1
Nov 25, 2021 · International Conference on Electronics, Computer and Computation
InDrive
InDrive is a global mobility platform best known for ride-hailing, connecting riders and drivers through a bid-based fare model.
Dec 2025 — PresentStandardized photo moderation rules and built vehicle image and document checks; tailored rules for Arabic-speaking markets raised auto-approval from 20% to 70%.
• Initiated cross-team alignment with business and moderation teams to establish unified rules for vehicle photo approval, replacing inconsistent team-by-team judgment with a standardized policy. • Developed a VLM-based vehicle recognition module using prompt engineering to extract license plate, brand, model, and color from vehicle photos, achieving 80%+ recognition accuracy. • Built an OCR pipeline to parse vehicle registration documents and cross-validate extracted data against vehicle photo rules, improving data consistency across the onboarding pipeline.
QazCode
QazCode is Beeline Kazakhstan's technology company, building digital products and services for the telecom operator and its customers.
Jul 2025 — Nov 2025Benchmarked generative image models and shipped content generation and document processing tools, including a CLIP-based classifier for new document types.
• Led research and benchmarking of 10+ generative AI models for image generation/editing, creating an evaluation framework used by the engineering team for model selection decisions. • Developed and deployed a GenAI content generation service for the marketing team using Gradio and Diffusers, enabling non-technical users to create AI-generated creative assets through an intuitive interface. • Designed a document image quality assessment solution for a biometric verification system, optimizing server-edge data flow and reducing processing overhead.
Kaspi.kz
Kaspi.kz is a major Kazakhstan technology and financial services ecosystem, spanning payments, banking and a consumer marketplace.
Sep 2023 — Jul 2025Built biometric and deepfake detection systems handling over 30 million photos monthly, with reported fraud savings up to 50 million KZT, and led a four-person team.
• Researched and developed a DeepFake detection model to identify various deepfake attacks. Created 5+ TB of deepfake datasets for training. Achieved 99% service accuracy and nearly 100% fraud coverage. Implemented a back-end service for the model using FastAPI, ONNX, and Docker, allowing the detector to process more than 30 million photos per month. • Researched the characteristics of virtual camera attacks and developed decision-making strategies against DeepFake attacks. These strategies influenced over 30 business processes and saved up to 50 million KZT in potential fraud damages.
CEREBRA.AI
Cerebra.AI is a Kazakhstan health technology company developing AI tools for brain imaging and stroke diagnosis for clinical use.
Sep 2022 — Mar 2023Developed MRI and CT image segmentation models and image processing tools for stroke diagnosis, plus a dashboard for weekly diagnosis statistics.
• Trained a segmentation model of Atlas of ASPECTS for MRI images using PyTorch Lightning. Performed hyperparameter tuning with Optuna. Article published on Medium. • Implemented a service with OpenCV to process MRI images and return coordinates of segmented brain areas. The service helped doctors detect damaged areas of the brain more efficiently. • Trained a segmentation model for hemorrhagic stroke blood classification in CT images using the 2.5D technique. This approach reduced the model’s size on the server while maintaining quality comparable to 3D models.
ML community of KZ
May 2020 — PresentFounded a Telegram channel for young professionals that grew to 4,000+ subscribers; organized 10 meetups and secured partnerships with four organizations.
• Created a platform for young professionals to share IT news and opportunities. 4000+ subscribers. • Organized 10 meetups with IT professionals, each attracting 200-400 attendees. • Secured official partnerships with JetBrains, Beetech Conf 2024, AppSecFest 2024 and CodeTalks 2024.
SDU University
2019 — 2023