Senior C++ Developer

На данный момент C++ Developer.

Мои компетенции и опыт

About

I'm a C++ developer with 5 years of experience building low-latency, high-load applications and production-grade C++ components for desktop and server platforms.

My core strength is modern C++ engineering: multithreading, asynchronous programming, memory-aware design, performance profiling, latency optimization, testing, and integration of complex components into reliable production systems. I worked on a large-scale legal document search engine serving 343 million documents, where performance, robustness, and maintainability were critical.

In addition to backend and systems development, I have practical experience in Machine Learning and Computer Vision pipelines: image and video preprocessing, OpenCV-based frame processing, object detection workflows, model training in Python/PyTorch, ONNX export, C++ inference integration, and runtime optimization. I’m comfortable working at the boundary between ML prototypes and production C++ code: taking a trained model, preparing inference artifacts, integrating it into an existing C++ pipeline, profiling the bottlenecks, and optimizing the full execution path.

I also participated in the Telegram ML Competition 2023, where I worked on a local code-snippet language classifier. The project involved training neural classification models, exporting them to ONNX, wrapping inference into a C++ shared library, minimizing dependencies, and optimizing execution under strict latency constraints.

I’m especially interested in real-time Computer Vision systems: object detection, object tracking, ball tracking, sports analytics, video processing, camera-control logic, and high-performance inference pipelines in C++.

Relevant Experience

C++ / Production Systems

Developed production-grade C++ components for high-load and latency-sensitive systems.

Worked on a legal document search engine processing hundreds of millions of documents.

Designed and implemented new components from scratch and integrated them into existing systems.

Optimized performance through profiling, algorithmic improvements, memory usage analysis, and refactoring.

Built asynchronous and multithreaded components using modern C++ and Boost.

Worked with networking, WebSockets, REST APIs, RPC, JSON serialization, and binary protocols.

Developed unit tests, benchmarks, and CI/CD-compatible workflows.

Participated in code reviews, debugging, architecture planning, and legacy code modernization.

Machine Learning / Computer Vision / Inference

Practical experience with Python/PyTorch model training, validation, preprocessing, and inference evaluation.

Experience preparing datasets, training classification and detection-style models, and evaluating accuracy/latency trade-offs.

Hands-on experience exporting trained models to ONNX and integrating inference into C++ applications.

Experience with C++ inference optimization: reducing allocations, minimizing preprocessing overhead, batching where appropriate, profiling the full inference path, and optimizing latency-critical code.

Familiar with ONNX Runtime and OpenVINO-style deployment workflows for CPU inference.

Experience with OpenCV-based image and video processing: frame capture, resizing, normalization, color conversion, ROI extraction, frame-by-frame processing, and preparing tensors for neural network inference.

Understanding of YOLO-style object detection pipelines: preprocessing, model inference, bounding-box decoding, confidence filtering, NMS, and postprocessing.

Understanding of object tracking pipelines: detector-tracker architecture, Kalman filters, IoU-based matching, data association, track lifecycle, target switching, and handling missed detections.

Interested in sports video analytics, ball tracking, player tracking, and PTZ camera-control systems.

Telegram ML Competition нужен доступ к резюме нужен доступ к резюме )

Participated in Telegram ML Competition 2023, focused on building a local library for detecting the programming or markup language of code snippets from Telegram messages.

Key work included:

Training neural classification models for noisy real-world code snippets.

Preparing and cleaning training data from public sources.

Designing a fast preprocessing and inference pipeline.

Exporting trained models to ONNX for local execution.

Integrating model inference into a C++ shared library.

Building the project with CMake for Linux/x86-64 deployment.

Optimizing inference latency under strict runtime constraints.

Reducing external dependencies and making the solution suitable for clean-system deployment.

Skills

C++

Modern C++11/14/17/20, STL, нужен доступ к резюме , нужен доступ к резюме , multithreading, concurrency, asynchronous programming, memory optimization, profiling, low-latency design, high-load systems, CMake, Linux development.

Computer Vision / ML

OpenCV, image processing, video processing, PyTorch, YOLO-style object detection, ONNX, ONNX Runtime, OpenVINO, model export, model inference, preprocessing, postprocessing, object tracking, Kalman filters, data association, ball tracking, sports analytics.

Backend / Infrastructure

gRPC, Protobuf, WebSockets, REST API, JSON, RPC, OpenSSL, rapidjson, glaze, fmt, pybind11, GoogleTest, Google Benchmark, Docker, Git, SVN, TeamCity, Jenkins, Grafana, SQL, MySQL, NoSQL, Redis.

Education

Master of Science in Laser Physics
Specialization: Optical Information Processing

Languages

Russian — native
English — professional working proficiency
German — elementary proficiency