Kuanysh Murat

Data Engineer

NITEC

Government-source data integration and Lakehouse services, Big Data / Smart Data Governance.

Builds data pipelines and ClickHouse models, pilots CDC, and develops internal AI data tools and ETL monitoring for a government data integration platform.

• Develop Python/SQL Airflow pipelines from PostgreSQL, MySQL, MS SQL Server and Oracle; adapt loads to schema changes and migrations and replace oversized full loads with incremental queries or CDC. Built a Jinja-based DAG generator from PostgreSQL metadata. • Implemented a Debezium/Kafka CDC pilot with bounded Airflow workers, offsets committed after successful inserts and retry-safe ClickHouse views. Write dbt/SQL joins and aggregations; model ClickHouse tables, sorting keys, partitions and projections.

Backend Developer

SANA

Kotlin school LMS using REST/gRPC, Hibernate/JPA and PostgreSQL.

Delivered Kotlin LMS import and access-control features, reduced email lookups from N database round trips to one, and wrote 20 tests.

• Delivered student/teacher imports across Protobuf contracts, HTTP annotations and router/service/repository layers. Built an Apache POI Excel reader with validation and precise row/column errors; enforced file ownership and role-based access. • Eliminated N+1 email lookups using one JPQL query and a hash set; reduced this stage from N database round trips to one. Added per-row results and handled transaction rollback on failure.

Machine Learning Junior Engineer

Yurt Tech

Python video-analysis services with Flask, RabbitMQ/aio-pika and OpenCV.

Built video analysis and inference services, raising fall detection accuracy from 62% to 93%, speeding inference over 25 times, and reducing irrelevant video traffic by 40%.

• Generated 100,420 training examples by projecting Kinect 3D skeletons into 2D. Compared models using cross-validation and built an RNN service on YOLO keypoints, improving fall-detection accuracy from 62% to 93%. • Accelerated YOLO/InternVL inference by over 25× on Intel Xeon using OpenVINO and quantization. Built a human-detection filter that cut irrelevant video traffic by 40% on average. • Implemented InternVL scene classification and used YOLO with OCR to label approximately 10,000 license plates, comparing EasyOCR, Tesseract and Qwen for quality and cost.

Research Assistant

Nazarbayev University

Linear-separability testing and distance approximation; three-student research team.

Implemented computational geometry and CUDA experiments for a research manuscript, automated benchmarks, and built a dataset service.

• Implemented C++ computational-geometry algorithms and synthetic-data generators; parallelized distance evaluation in CUDA. Contributed a 2D-to-3D theoretical extension, proofs and experiments to the team’s research manuscript. • Automated repeated benchmarks with Python subprocess limits and timeouts; compared correctness, runtime and SVM baselines using pandas, Matplotlib and seaborn. Built a FastAPI/MongoDB dataset service served through Uvicorn and Nginx.

C++/OOP/DSA instructor

Volunteer teaching

Volunteered to teach C++, OOP and data structures to nine students, creating lessons, exams and contests and giving project feedback.

• Volunteered as a C++/OOP/DSA instructor for nine students, designing lessons, exams, contests and project feedback.

Bachelor, Computer Science

Nazarbayev University