Senior Computer Vision Engineer (Scientific Imaging)

We are looking for a Senior Computer Vision Engineer to lead the two-person core computer-vision team of an X-ray inspection proof of concept for wind turbine blades: six to eight months with a path to a product. Around you: a technical lead, a project manager, shared QA, an NDT expert and a blade engineer part-time, and the acquisition partner's engineers.

• The input is a single stitched X-ray image of a whole blade, tens of thousands of pixels long, delivered by the acquisition partner. Reconstruction from raw detector frames is outside the role and outside the initial scope. The first task is to find a thin, dense metal conductor inside the composite, follow it from one end of the blade to the other, and tell where it is broken. The second task is a feasibility question: evaluate whether missing adhesive or an open separation in a joint deep inside the structure produces a detectable and repeatable indication, and at what size. We expect classical image processing to be the starting point: normalising brightness across material that changes thickness by an order of magnitude, handling stitching seams, tracing a linear structure through clutter, calibrating a threshold on physical reference samples. Learned components are used only where experiments show they add value.
• You do not need X-ray or non-destructive-testing experience. We teach the physics and pay for the reading time. What we need is someone who has entered an unfamiliar imaging domain before and got to a working method fast, and who can tell the difference between "the algorithm missed it" and "there is nothing in the image to find". One honest outcome of the first phase is that a defect type cannot be seen reliably; we want the person who can show why.

Responsibilities:

• Assess the first images: what is visible, at what contrast, where the method will and will not work

• Define image-quality requirements the acquisition partner must meet, and check every delivery against them

• Build preprocessing: tiling, thickness normalisation, denoising, seam handling

• Build the conductor tracing: linear-structure filters, trace graph, branches, rejection of look-alike structures

• Run the feasibility gate for the second defect type on a simulator and on reference samples with known defects; the outcome may be a documented no-go

• Own the method: the physical and algorithmic hypothesis, pipeline architecture, thresholds and the policy for uncertain cases, validation design, the technical decision on every miss, and technical sign-off

• Review the second engineer's implementation and keep the end-to-end pipeline reproducible

• Work with the acquisition partner's engineers, together with the technical lead, on image format, calibration, metadata and stitching problems

• Tune the pipeline on the first images from an installed blade and report what changed and why

• Work with the NDT expert and the blade engineer on thresholds and on every miss

• Be on site at one test campaign at a wind farm (about one week) to check images before the equipment leaves

Requirements:

• 4+ years in computer vision or image processing on real, imperfect images

• Classical image processing at a working level: filtering, ridge and line detection, morphology, segmentation, connected components, graph-based tracing

• Experience with images too large for memory: tiling, overlap, reconstructing coordinates in the full image

• Strong Python with NumPy, OpenCV, scikit-image or equivalents

• At least one project where the data was scarce and you had to design how it was labelled and validated

• Experience quantifying image quality (contrast-to-noise, resolution, blur, repeatability) and designing controlled experiments that separate an imaging limitation from an algorithm limitation; any modality

• At least one project where you moved into a modality you did not know before (medical, microscopy, satellite, industrial, scientific) and can explain how you got up to speed

• English good enough to read papers and talk to the client's engineers

Nice to have:

• Radiography, CT, NDT, composites, weld or semiconductor inspection

• Physics or engineering background: attenuation, scatter, beam hardening, geometric unsharpness, detector correction

• Simulation of imaging (any X-ray or optical simulator)

• Small-dataset deep learning: segmentation, transfer learning, hard-negative mining

• Experience influencing how data is physically acquired, not only processing what you were given

• Use of AI coding agents for prototyping and tooling