We’re looking for vision engineers who know that not every problem needs a neural network — and that some problems can’t be solved without one. You’ll work across the full spectrum of computer vision: classical geometric methods like camera calibration, stereo, and multi-view geometry, alongside modern deep learning for detection, segmentation, and tracking. Knowing which tool fits which problem is exactly the judgment we’re hiring for.

Our vision projects are not academic exercises. They run on factory floors, in vehicles, and on edge devices with real latency and accuracy constraints — from quality inspection in manufacturing through autonomous driving perception to enterprise-scale image and video analytics.

What We Offer

AI Grant — Stop talking about AI and start building it. Our AI Grant gives you dedicated budget and resources to turn your wildest AI idea into a working project, backed by two paid weeks to focus on nothing else.

AI Center of Excellence — Work alongside specialists in agentic AI, sovereign AI, generative and discriminative AI. This isn’t a siloed team — it’s the people you’ll learn from and build with daily.

Your tools, your choice — Full access to AI-powered development tools including Claude, Cursor, and GitHub Copilot. Pick what works best for you.

Real project variety — From generative AI for legal document compliance through agentic systems in manufacturing environments to enterprise-scale AI platforms, computer vision, and autonomous driving. You won’t get bored.

Conference and speaking support — Want to attend conferences? We’ll back you. Want to speak at them? Even better — we’ll support you with dedicated preparation time and bonuses.

Your tasks

Design and implement vision systems combining classical and learning-based approaches — choosing the right method for the constraints of each problem
Apply geometric computer vision in practice: camera calibration (intrinsic/extrinsic), stereo and multi-view geometry, feature detection and matching, pose estimation, structure-from-motion, and visual SLAM
Build, train, and fine-tune deep learning models for object detection, semantic and instance segmentation, multi-object tracking, OCR, and anomaly detection using PyTorch
Develop image processing pipelines with OpenCV and NumPy: filtering, morphology, thresholding, contour analysis, and color space operations
Optimize models and pipelines for deployment: quantization, pruning, ONNX/TensorRT conversion, and real-time inference on edge devices and GPUs
Own data quality end-to-end: dataset design, annotation strategy, augmentation, and systematic error analysis on real-world imagery
Build evaluation frameworks that measure what matters in production — not just mAP on a benchmark, but failure modes under real conditions
Collaborate with AI Architects and ML engineers to integrate vision components into larger systems, on-prem and in the cloud

Requirements

At least 5 years in computer vision engineering, with vision systems shipped to production — not just research prototypes
Solid grounding in geometric vision: projective geometry, camera models, epipolar geometry, and 3D reconstruction fundamentals
Hands-on deep learning experience: training and deploying CNN- and transformer-based models (e.g., YOLO family, Mask R-CNN, DETR, SAM) for real applications
Strong Python and OpenCV skills; comfort with PyTorch and the surrounding ecosystem (torchvision, Albumentations, ONNX)
Experience optimizing inference for constrained environments — latency budgets, embedded hardware, or high-throughput video
The judgment to pick a homography over a transformer when the problem calls for it — and to defend that choice
Ability to work autonomously while collaborating effectively with architects, engineers, and product teams
Fluent English (both written and spoken)
Fluent Polish required
Residing in Poland required

Nice to have

C++ skills for performance-critical pipelines and integration with existing vision systems
Previous work with visual SLAM, sensor fusion (LiDAR, radar, IMU), or autonomous driving perception stacks
Exposure to vision-language models (CLIP, Grounding DINO) and foundation-model-based labeling or zero-shot pipelines
Experience deploying on NVIDIA Jetson, industrial cameras (GenICam/GigE), or cloud vision services on Azure, AWS, or GCP

Job no. 260612-LXSXE

Sii ensures that all hiring decisions are made solely on the basis of qualifications and competence. We are committed to equal and fair treatment of all, regardless of legally protected characteristics. At Sii, we promote a diverse and inclusive work environment, in full compliance with applicable anti-discrimination laws.

Benefits For You