Skip to content

Computer Vision Engineer (m/f/d) – Embedded / Eye Tracking

Metzingen, Germany

Apply Now!

At ZensTrack, we develop Merlin — a high-precision eye and head tracking system for research, medical technology, and sports performance analysis.

You will work on the complete vision stack: from embedded cameras and real-time image processing to deep-learning eye-feature detection, 3D gaze estimation and sensor fusion. Our systems run on resource-constrained hardware (Raspberry Pi CM4 + Hailo-8 NPU) at frame rates up to 500 Hz and are used in real-world applications worldwide.

We are looking for someone who wants to take technical ownership and enjoys solving challenging computer vision problems pragmatically and cleanly within a small engineering team.

Responsibilities

  • Develop and improve our DNN-based eye-tracking pipeline (pupil, iris, eyelid, eyeball, gaze, iris torsion)
  • Train, optimize and deploy neural networks for real-time inference on embedded hardware (ONNX → Hailo-8, INT8 quantization)
  • Real-time image processing; optimize performance for high frame rates and low latency
  • Work on 3D model-based gaze / eyeball estimation and IMU sensor fusion
  • Develop calibration and camera-control algorithms (autofocus, exposure)
  • Contribute to and extend our C++ vision stack

Must-have

  • 3+ years of hands-on experience with Computer Vision / Image Processing
  • Solid C++ skills
  • Deep learning for vision tasks — training and evaluating CNNs, especially semantic segmentation (e.g. U-Net)
  • Experience deploying models on resource-constrained / edge hardware (quantization, ONNX, NPU/GPU or similar)
  • Experience with OpenCV or similar frameworks
  • Structured and reliable engineering mindset; interest in embedded and real-time systems
  • Fluent written and spoken English; basic German;
  • Valid work authorization for Germany required

Nice-to-have

  • Eye tracking, gaze estimation, 3D eye-model fitting, ophthalmic or medical imaging
  • Edge-AI deployment: Hailo (Dataflow Compiler), NVIDIA Jetson, Coral/Edge-TPU or comparable NPUs
  • Model optimization: quantization, pruning, knowledge distillation, neural architecture search
  • Embedded Linux experience; ARM cross-compilation; SIMD / GPU / NEON optimization
  • IMU fusion, signal processing, camera calibration
  • Experience integrating open-source eye-tracking stacks (e.g. OpenIris, pye3d, RITnet/EllSeg); foundation-model-assisted annotation (e.g. SAM)
  • Experience with or interest in Rust
  • Experience in medical or regulated environments; research background or publications

Why this role is interesting

  • Work on a complete real-time vision system instead of a narrowly scoped subsystem
  • Solve challenging engineering problems at the intersection of computer vision, deep learning, embedded systems, optics, and sensor fusion
  • See your work directly in products used in research, medical technology, and sports performance analysis
  • High technical ownership and direct influence on product and architecture decisions
  • Small engineering team with short feedback loops and fast iteration cycles
  • Build systems that operate on resource-constrained hardware at frame rates up to 500 Hz
  • Engineering-focused culture with an emphasis on clean, reliable, and maintainable systems
Eye Tracking Platform

See the system

MERLIN

What you can expect

  • Responsibility for a complete real-time vision system — not just a small subsystem
  • Direct influence on product direction and technical architecture
  • Challenging problems across computer vision, embedded deep learning, optics, and sensor fusion
  • Small engineering team with short decision paths and high technical ownership
  • Products used in real research, medical, and sports environments worldwide
  • An engineering culture focused on clean, reliable, and maintainable systems
  • Flexible working hours and hybrid work options by arrangement
  • Modern development hardware and tools



Apply Now!