Job Description: We are a European HealthTech startup operating in stealth mode, developing a clinical diagnostic support tool. We are looking for an experienced Computer Vision / Machine Learning Engineer to build an automated image analysis pipeline for dental radiography. The core of the project involves processing 2D radiographic images to automatically detect specific anatomical landmarks with extremely high precision. Based on these detected landmarks, the algorithm must calculate predefined geometric parameters (distances, angles, and ratios). In a later phase, the project will expand to incorporate 3D data, including STL files and Conebeam CT-scans. Key Responsibilities (Phase 1): Develop and train a robust Deep Learning model (e.g., coordinate regression, heatmap-based approaches like U-Net or HRNet) for keypoint/landmark detection on 2D medical X-rays. Build a mathematical pipeline to automatically calculate specific diagnostic metrics based on the detected coordinates. Ensure the model performs reliably across varying image qualities and contrasts. Required Expertise: Proven track record in Computer Vision and Deep Learning (PyTorch or TensorFlow). Specific experience with Medical Image Processing (e.g., DICOM formats, X-ray enhancement). Strong understanding of Landmark Detection / Keypoint Estimation algorithms. Proficiency in OpenCV and Python. (Bonus) Experience with 3D point clouds, meshes, and STL processing for future phases. Confidentiality: Due to the proprietary nature of this project, specific diagnostic methodologies (such as the exact cephalometric analyses used) will only be shared with shortlisted candidates after signing an NDA. To Apply, please include: A brief overview of your experience specifically related to medical imaging or landmark detection. (General AI or LLM/ChatGPT wrappers are not relevant for this job). Your answer to the screening question below. Screening Question (Required): Medical X-rays often suffer from low contrast and overlapping anatomical structures. Which specific Deep Learning architecture or technique would you recommend for high-precision landmark detection in such conditions, and why?