Senior AI Engineer- Imaging

Roche•South San Francisco, CA

5d•$168,100 - $312,300•Hybrid

About The Position

As an imaging-focused machine learning engineer on the AI Enablement team, you will collaborate with computational scientists, biologists, bench scientists, AI/ML experts, product leaders, software engineers, and other partners. The work bridges scientific intuition with quantitative analysis, enables richer phenotypic analysis, and provides multimodal capabilities across large-scale scientific datasets. You will build, own, and improve scalable AI/ML and computer vision systems for scientific imaging data, including microscopy, digital pathology, and biomedical imaging, to accelerate discovery and development of treatments. The role involves developing scientific web applications, backend pipelines, computer vision models, evaluation tools, and cloud infrastructure for high-throughput imaging data. You will partner with scientific and product collaborators to translate workflow needs into technical specifications, optimize model and pipeline performance, contribute to architecture and MLOps practices, and work across the stack as needed. The position is expected to be onsite at the South San Francisco campus at least three days per week. Relocation assistance is not available. The expected annual salary range for the primary San Francisco, California location is $168,100–$312,300. The role is open at SE6 and SE7 levels, with final leveling determined after interviews. A discretionary annual bonus may be available based on individual and Company performance.

Requirements

PhD, or a Master’s degree with 3–5 years of experience, or a Bachelor’s degree with 4–7 years of experience, in Computer Science, Biomedical Engineering, Electrical Engineering, Computational Biology, or a related technical field focused on machine learning engineering and computer vision.
Strong proficiency with AI/ML frameworks and computer vision libraries such as PyTorch, TensorFlow, OpenCV, MONAI, Scikit-Image, torchvision, Napari, or Bio-Formats.
Deep expertise in computer vision algorithms, such as CNNs, Vision Transformers, UNet, SAM/Segment Anything, and Cellpose.
Expert knowledge of machine learning theory, image processing fundamentals, object detection, instance segmentation, and quantitative image analysis.
Demonstrated experience with high-dimensional scientific imaging data, such as fluorescence or brightfield microscopy, digital pathology whole slide images, high-content screening, or multiplexed imaging.
Strong knowledge of GPU acceleration, parallel computing, CUDA, and ML performance optimization for large image datasets.
Experience with Kubernetes, relational and NoSQL databases, object storage, and cloud-native architectures in public clouds; AWS is preferred.
Proven understanding and application of software engineering best practices, CI/CD, and MLOps principles.
Excellent communication skills and ability to build trusted partnerships with scientific collaborators, bioimage analysts, and cross-functional teams.
Ability to quickly acquire new technologies, domain knowledge, and programming languages, with a passion for continuous learning.

Nice To Haves

Experience with multimodal data fusion, combining scientific imaging with genomics, transcriptomics, or clinical tabular data.
Experience working directly with biomedical scientists or in a pharma/biotech research environment.
Experience with workflow automation, GenAI, Vision-Language Models, and/or AI agents.

Responsibilities

Design, develop, and test robust, scalable, maintainable AI/ML scientific web applications and backend pipelines for high-throughput scientific imaging processing and visualization.
Build, evaluate, and fine-tune computer vision and deep learning models for tasks such as cell segmentation and classification, phenotypic profiling, tissue segmentation, feature extraction, and image registration.
Build tools to evaluate AI/ML and vision model performance, establish benchmark datasets, and quantify model quality, robustness, and drift over time.
Partner with product managers, bioimage analysts, and research scientists to understand workflows, shape requirements, and translate scientific imaging needs into technical specifications.
Develop and maintain scalable cloud infrastructure and data pipelines for collecting, structuring, storing, and accessing large-scale multidimensional imaging datasets.
Optimize GPU memory footprint, batching, and inference throughput for large image pipelines and deep neural network models.
Contribute to architectural decisions, code reviews, and the evolution of engineering and MLOps processes for computer vision models.
Contribute across the full stack and outside core areas of expertise as needed.
Stay current with emerging computer vision technologies and industry best practices, and promote continuous learning, collaboration, and curiosity.