Principal Computer Vision Scientist

About OrthoAIOrthoAI is building AI-driven decision intelligence for dentistry. Our initial focus is turning multimodal clinical information into structured, explainable support for high-friction authorization workflows. We are building from Abu Dhabi for the UAE and wider GCC, with clinical validity, traceability, and user safety designed into the product from the start.The roleTurn heterogeneous medical imagery and clinical inputs into calibrated, structured features that support downstream clinical reasoning and authorization workflows. You will own rigorous computer-vision research, reproducible evaluation, clinically meaningful error analysis, and controlled transfer into production.What success looks like•    Within 90 days, reproduce and audit the current baseline; define task formulations, leakage controls, evaluation slices, uncertainty methods, and experiment standards.•    Within six months, deliver a validated multimodal perception baseline with calibrated confidence, ablation evidence, and clinician-readable failure analysis, with SOTAs methodolgical contributions.•    Within 12 months, improve robustness across available sites and acquisition conditions, add distribution-shift checks, and package a controlled production contract.•    Create reusable datasets, evaluation harnesses, model cards, and decision records that make positive and negative results legible.Key responsibilities•    Develop models for relevant combinations of segmentation, detection, landmarks, geometric measurement, 3D vision, classification, and multimodal fusion.•    Design datasets and evaluation protocols that prevent leakage and expose clinically material errors.•    Measure calibration, selective prediction, abstention, and robustness under realistic distribution shifts.•    Partner with clinicians on ontology, ground truth, error severity, and clinically meaningful acceptance criteria.•    Partner with engineering on preprocessing contracts, reproducible packaging, latency/cost constraints, and monitoring.•    Track relevant research and help decide what to publish, patent, retain as know-how, or stop pursuing.Required qualificationsEducation or equivalent evidencePhD in computer vision, machine learning, medical imaging, applied mathematics, or a closely related field; or an equivalent research record demonstrated through reproducible work, publications, patents, or deployed systems.Experience expected•    Typically 5+ years of applied computer-vision or machine-learning research, including doctoral research where relevant.•    Demonstrated ownership of a research problem from dataset and formulation through evaluation, error analysis, and handoff.•    Experience with real-world image variation, imperfect labels, distribution shift, class imbalances, and clinically or operationally meaningful failure modes.•    Evidence of collaboration with domain experts and engineers, not only independent model development.Technical and functional capability•    PhD in computer vision, machine learning, medical imaging, or equivalent demonstrated research depth.•    Strong Python and modern deep-learning practice with code another researcher can reproduce and extend.•    Depth in at least two relevant areas: segmentation, detection, landmarks, 3D vision, multimodal fusion, uncertainty, domain adaptation, or weak/self-supervision.•    Rigorous dataset design, calibration, leakage prevention, ablation, and failure analysis.•    Ability to translate an ambiguous clinical construct into a testable modeling and evaluation problem.Preferred qualifications (Plus to have)•    Dental, orthodontic, craniofacial, radiology, or adjacent medical-imaging experience.•    Panoramic/cephalometric imaging, CBCT, landmark measurement, or cross-modal clinical representation.•    Research-to-production transfer in a regulated or high-consequence environment.Evidence that will stand out•    One reproducible research artifact: formulation, dataset, baselines, ablations, failures, and what did not work.•    A design for evaluating multimodal perception across sites, devices, or acquisition conditions.•    Code, models, datasets, patents, or publications that another expert can inspect and challenge.How we work at OrthoAI•    Outcome ownership: one accountable owner, clear interfaces, and visible follow-through.•    Believability by domain: ideas are weighted by relevant demonstrated success and causal reasoning, not title.•    Written learning: important decisions, misses, root causes, and machine changes are recorded.•    Cross-disciplinary rigor: clinical, scientific, engineering, commercial, and operational evidence must connect.•    Human accountability: AI supports decisions; accountable people make and own them.Our values are a gateResults do not compensate for a values breach. We look for demonstrated behavior consistent with:•    Truth over comfort: surface inconvenient facts and change your view when evidence changes.•    Ownership without excuse: take outcomes end to end, including dependencies and recovery.•    Evidence over opinion: distinguish observation, inference, and hypothesis.•    Patient and user safety is inviolable: stop and escalate when a safety boundary is at risk.•    Speed with rigor: shorten feedback loops without removing the controls the risk requires.What we offer•    Founding-stage scope with direct ownership of a company-critical outcome.•    Close collaboration with the CEO and the clinical, scientific, engineering, and commercial founding team.•    Meaningful influence over systems, standards, hiring, and the roadmap within your domain.•    Technically and operationally serious work using real healthcare workflows and carefully governed data.•    A premium compensation package positioned above market, including a competitive base salary, meaningful Employee Stock Ownership Plan (ESOP) equity, and comprehensive benefits — confirmed in full for shortlisted candidates at offer stage.Selection process•    LinkedIn application and evidence screen against the role outcomes.•    Structured interviews focused on demonstrated decisions, results, and lessons.•    A paid, role-specific working session based on a realistic OrthoAI problem.•    Values-gate interview and candidate-authorized references.Location and eligibilityThis is a full-time, Hybrid role based in Abu Dhabi, UAE. Candidates must be able and willing to work in Abu Dhabi. Please state your current UAE work authorization or the support you would require; this information is assessed only where job-related and does not replace evaluation of role capability.Apply through LinkedInUse the LinkedIn application for this role. Include your CV and a concise example of a problem you owned end to end: the outcome, your decisions, the evidence you used, what went wrong, and what changed because of your work. Do not include patient data, unnecessary personal information, or confidential employer material.