Beograd
01.10.2026.
About the role
Trego is looking for an experienced Computer Vision Algorithms Engineer to take ownership of computer vision models for smart occupancy-sensing applications.
You will design, train, evaluate, optimize, and deploy models for privacy-preserving occupancy sensors. This is a hands-on role combining applied research with production engineering, with a strong focus on reliable performance on resource-constrained edge devices.
What you will do
- Develop and maintain models for people detection, counting, and tracking, combining deep learning with classical computer vision where appropriate.
- Build reproducible data, training, evaluation, and model-release pipelines.
- Define datasets, annotation requirements, and evaluation protocols that reflect real deployment conditions.
- Analyze field data and model failures to improve accuracy and overall product performance.
- Optimize and deploy models within accuracy, latency, memory, compute, and power constraints.
- Collaborate with embedded software, hardware, QA, product, and customer-facing teams on integration, validation, and issue investigation.
What you bring
- At least 5 years of industry experience in computer vision or deep learning, including object detection and tracking.
- Strong proficiency in Python and C++, together with NumPy, OpenCV, and PyTorch or TensorFlow.
- Demonstrated experience taking computer vision models from research into production.
- Hands-on experience optimizing neural networks for constrained edge hardware, including INT8 quantization and performance profiling.
- A strong understanding of model evaluation, dataset quality, overfitting, distribution shift, and failure analysis.
- An M.Sc. in Computer Science or Electrical Engineering, with a specialization in computer vision.
- The ability to work independently and communicate clearly with technical and nontechnical stakeholders.
- Professional working proficiency in English.
Additional advantages
- Experience with YOLO models and model-conversion tools such as ONNX or TensorFlow Lite.
- Experience with embedded Linux, camera pipelines, real-time systems, or NPU and DSP toolchains.
- Experience with low-resolution or wide-angle imagery, people counting, or occupancy analytics.
- Experience with MLOps.
- Experience using AI-assisted development tools.
Upoznaj kompaniju
Trego (www.trego-usa.com) is a high-end provider of turn-key solutions and RT Embedded SW.
Our Main Expertise are Embedded Linux, Video-based products, and Semiconductors.
Trego's Engineers are exposed to diverse state-of-the-art technologies (Architectures, tools, environments etc.).
They perform mostly proper R&D tasks and hardly any maintenance.
We are a small group of experts, working in a flexible, tolerant and purely professional atmosphere.
Preporuke se učitavaju...
