JOB DESCRIPTION
Senior Data Scientist – Medical Imaging AI
Machine learning engineering and computer vision for clinical imaging
LOCATION
Gurugram, India
WORK MODEL
On-site 5 days/week
EXPERIENCE
3–5 years
About ImagingIQ
ImagingIQ is building AI-assisted medical-imaging and radiology workflow solutions that help clinical teams analyse studies, review structured findings and integrate AI outputs into real-world reporting workflows. Our current work centres on brain MRI oncology and is expanding into brain CT, trauma, vascular, oncology and volumetric applications. Our systems work with radiologists in the loop: AI outputs support clinical review and do not replace the final clinical decision.
The role
We are looking for a hands-on Senior Data Scientist who operates at the intersection of machine learning engineering, computer vision and medical imaging. This is not a business analytics or dashboarding role. You will take ownership of clinically relevant imaging problems from data definition and experimentation through validation, optimisation, deployment and post-deployment improvement.
You will work with multi-sequence MRI and CT data, radiologists, data annotators, product and platform engineers. The role requires someone who can reason about image quality, anatomy, spatial correctness and dataset bias while also writing production-quality code and helping models run reliably within an integrated medical-imaging pipeline.
Core mandate: Build medical-imaging AI that is technically rigorous, clinically interpretable, reproducible and deployable—not merely accurate on a benchmark dataset.
What you will own
- Develop and improve 2D and 3D computer-vision models for medical-image segmentation, classification, detection, localisation, measurement and related clinical decision-support tasks.
- Own substantial model-development workstreams end to end: problem formulation, dataset design, preprocessing, experimentation, evaluation, optimisation, release support and production monitoring.
- Design robust preprocessing pipelines for DICOM and NIfTI data, including series selection, orientation handling, resampling, registration, bias correction, skull stripping, intensity normalisation and quality checks where applicable.
- Translate segmentation and classification outputs into clinically meaningful measurements and structured results such as tumour volumes, anatomical localisation, volumetric comparisons and other quantitative imaging features.
- Work closely with radiologists and clinical reviewers to define acceptance criteria, analyse failure modes, incorporate structured feedback and distinguish clinically meaningful errors from metric-only differences.
- Build representative training and validation datasets; identify leakage, label noise, class imbalance, acquisition bias and site/vendor/protocol shifts; and define appropriate patient-level and centre-level splits.
- Evaluate models beyond aggregate accuracy using task-appropriate metrics, subgroup analysis, calibration, uncertainty, error taxonomies and clinically relevant operating points.
- Optimise and productionise models using formats and runtimes such as ONNX, NVIDIA Triton, TensorRT, OpenVINO or ONNX Runtime, balancing quality, latency, throughput, memory and operational reliability.
- Collaborate with platform and DevOps engineers on APIs, containers, queues, cloud deployment, observability, versioning, rollback and reproducible release processes.
- Maintain clear technical documentation and traceability for datasets, experiments, model versions, known limitations, validation evidence and changes in a regulated product-development environment.
- Review code and experimental plans, mentor junior data scientists, raise technical standards and help break research or product goals into executable engineering work.
Required qualifications
- 3–5 years of relevant industry experience in machine learning, computer vision or medical-image analysis, with meaningful hands-on ownership of model development.
- Strong Python programming skills and practical experience with PyTorch or an equivalent deep-learning framework.
- Solid understanding of modern computer vision, including segmentation and classification, loss functions, augmentation, optimisation, validation design and systematic error analysis.
- Experience working with 3D imaging data and libraries such as MONAI, SimpleITK/ITK, NiBabel, pydicom, ANTs or comparable tools.
- Working knowledge of DICOM and NIfTI, image geometry, coordinate systems, spacing, orientation and transformations between image spaces.
- Experience evaluating models using metrics appropriate to imbalanced classification and segmentation problems, such as sensitivity, specificity, precision, recall, F1, ROC-AUC, Dice and surface-distance measures.
- Ability to write maintainable, testable code and work effectively with Git, Linux and containerised development environments.
- Ability to communicate assumptions, limitations, trade-offs and evidence clearly to engineering, product and clinical stakeholders.
- Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Data Science, Biomedical Engineering, Electronics, Mathematics or a related quantitative discipline, or equivalent relevant experience.
Preferred qualifications
- Experience with brain MRI, CT, radiology AI, multi-sequence imaging, anatomical segmentation, registration, atlas-based analysis or volumetry.
- Experience taking a model from research code into a production inference service, including model conversion, profiling, optimisation and monitoring.
- Exposure to PACS-integrated workflows, DICOM SEG, structured clinical outputs or medical-image viewers.
- Experience with AWS, ML experiment tracking, model registries, CI/CD or production MLOps practices.
- Experience working with multi-centre or real-world clinical data and collaborating directly with radiologists, clinicians or trained medical annotators.
- Awareness of design controls, risk management, traceability and documentation expectations for medical-device or other regulated software development.
- Relevant publications, patents, open-source contributions or demonstrated work in medical imaging or applied computer vision.
What success looks like
- Model changes are supported by reproducible evidence and improve real-world clinical acceptance rather than only internal benchmark scores.
- Datasets, experiments, releases and limitations are traceable, reviewable and understandable to other members of the team.
- Inference components are reliable enough to operate within production turnaround-time, concurrency and integration constraints.
- Radiologist feedback is converted into clear technical hypotheses, prioritised improvements and measurable validation plans.
- Junior team members receive practical technical guidance while you remain directly involved in implementation and problem solving.
Why this role matters
You will work on problems where model behaviour, image geometry, software reliability and clinical interpretation all matter simultaneously. The role offers substantial ownership over how medical-imaging AI is built, validated and delivered into radiology workflows, with the opportunity to influence both the technical architecture and the product’s real-world clinical usefulness.
Pay: ₹1,200,000.00 - ₹1,500,000.00 per year
Benefits:
Ability to commute/relocate:
- Gurugram, Haryana (Gurugram): Reliably commute or planning to relocate before starting work (Required)
Application Question(s):
- How many years of hands-on experience do you have in ML/computer vision, and how much of that is specifically in medical imaging?
- Have you worked with 2D or 3D medical-image data?
- Have you developed models for medical-image segmentation, classification, detection, localisation, or volumetry?
- Have you taken an ML/CV model from research or experimentation into production?
If yes, ask what they used for deployment—ONNX, TensorRT, NVIDIA Triton, ONNX Runtime, Docker, etc.
- Have you worked with clinical teams, radiologists, medical annotators, or healthcare professionals?
If yes, what was your role in incorporating their feedback into the model?
Work Location: In person
