What you'll do- Own a multimodal ML work-stream from problem definition through experimentation, evaluation, deployment, and iteration.
- Translate clinical and product needs into clear ML objectives, data strategies, model approaches, and success criteria.
- Build and evaluate modern ML systems, including transformers, self-supervised learning, weak supervision, detection, localization, and segmentation.
- Work with image, report, and other clinical data to develop systems that are useful in real radiology workflows.
- Design rigorous evaluations that go beyond aggregate offline metrics, including clinically meaningful operating points, robustness, calibration, and performance across relevant data slices.
- Partner with engineering to productionize models, make practical system tradeoffs, and learn from performance after launch.
- Investigate failure modes such as laterality errors, poor image or report grounding, hallucination, dataset bias, domain shift, and workflow disruption.
- Communicate research findings and technical decisions clearly through design documents, experiment reviews, and presentations to technical and clinical partners.
- Contribute to the research roadmap by identifying promising approaches, sharing learnings, and helping the team decide what to pursue next.
- Mentor less experienced researchers and engineers through project collaboration, code and experiment reviews, and technical guidance.
What we're looking for- Strong applied experience in computer vision, NLP, or deep learning, with a track record of independently designing experiments, analyzing results, and turning findings into working systems.
- Experience owning substantial ML projects across the full lifecycle, from data and modeling through production delivery.
- Deep hands-on ability in Python and PyTorch, with strong intuition for model architecture, data quality, experimentation, and evaluation.
- Experience with modern vision or multimodal techniques such as vision transformers, contrastive learning, masked image modeling, or weak supervision, etc.
- The judgment to connect model performance to real user and clinical outcomes, including knowing when a benchmark improvement is not enough.
- Strong collaboration skills across research, engineering, product, data, and clinical teams.
- Clear written and verbal communication, including the ability to explain technical tradeoffs to both ML experts and clinical partners.
- Typically 4+ years of relevant applied ML research or engineering experience, or equivalent scope and impact. We calibrate on demonstrated ownership rather than title or exact tenure.
- An MS, PhD, or equivalent practical experience in Computer Science, Electrical Engineering, Machine Learning, Biomedical Engineering, or a related quantitative field.
Nice to have- Experience with medical imaging, radiology, healthcare, or another high-stakes application area.
- Familiarity with chest X-ray, CT, MRI, mammography, or other clinical imaging modalities.
- Experience with DICOM, image-report pairing, medical data de-identification, radiology workflows, or clinically derived labels.
- Experience evaluating models across patients, sites, scanner vendors, protocols, or other sources of distribution shift.
- Familiarity with clinical validation, FDA or HIPAA considerations, or other regulated and privacy-sensitive environments.
- Experience with 3D vision, longitudinal imaging, report generation, or clinical decision support.
- Publications, open-source contributions, or other evidence of research credibility.
What success looks likeYou'll own and advance a meaningful research track from ideation through production. You'll establish a strong understanding of the clinical problem, build a credible data and evaluation strategy, deliver models that perform reliably in practice, and help the team learn from real-world use.
You'll also become a trusted technical partner to the researchers, engineers, product leaders, data teams, and clinicians working on the broader ML roadmap. Over time, you'll help raise the quality of research and technical decision-making through strong experimentation, clear communication, and thoughtful mentorship.
Our working styleWe're a remote-first company with a highly collaborative, mission-driven research and engineering culture. We value direct communication, intellectual honesty, strong ownership, and practical judgment. The best work here comes from people who can go deep technically, stay close to the clinical context, and make progress even when the problem and the path are not fully defined.
This role is U.S. remote, with San Francisco Bay Area preferred. We encourage people from a wide range of backgrounds to apply. If the scope of this role excites you but your experience does not match every bullet, we would still love to hear from you.
Location Details:For roles listed as San Francisco - Onsite:
- This role will be based in our San Francisco office and we expect employees to work onsite four days per week. The remaining time may be worked remotely or onsite, depending on team and business needs.
For roles listed as United States - Remote:
- This role is open to candidates located anywhere in the United States.
For roles listed as San Francisco - Onsite + United States - Remote:
- We will prioritize candidates who can work onsite four days per week in San Francisco, while also considering remote candidates located anywhere in the United States.
For US-Based Full-Time Roles, Rad AI offers a variety of benefits, including:- Comprehensive Medical, Dental, Vision & Life insurance
- HSA (with employer match), FSA, & DCFSA
- 401(k)
- 11 Paid Company Holidays
- Flexible PTO policy
- Annual company-wide offsite
- Periodic team offsites
- Annual equipment stipend
- For roles based outside the US, your recruiter can share more details
