Data Engineer with python and date

Senior Architect, Algorithm Development WorkbenchLocation: San Francisco Bay Area, CA

Workbench Architecture and Design

• Define the end-to-end architecture of an algorithm development workbench, covering environment provisioning, data access patterns, experiment lifecycle management, and compute resource orchestration
• Design secure, scoped data access mechanisms enabling data scientists to access versioned imaging datasets and annotations without violating data governance or PHI boundaries
• Architect experiment tracking and lineage capabilities, ensuring full reproducibility of model training runs from data version through hyperparameters to model artifact
• Define the promotion pathway from exploratory experimentation to regulatory-grade model validation, including audit trail and artifact management requirements

ML Platform Integration

• Integrate workbench tooling with underlying data lake and imaging platform layers, including versioned dataset APIs, annotation outputs, and study-level data packages
• Define compute environment standards for model training workloads, including GPU resource management, containerization, and dependency isolation
• Establish standards for model packaging, versioning, and handoff to downstream deployment or regulatory validation pipelines
• Collaborate with data architects to define the data access API surface exposed to workbench consumers

Regulated Delivery

• Ensure workbench architecture meets GxP and SaMD requirements for software used in algorithm development for clinical or regulatory purposes
• Design audit trail and logging capabilities covering data access, experiment execution, model versioning, and promotion decisions
• Support IQ/OQ/PQ validation activities and produce required architecture documentation for Computer Software Assurance (CSA) qualification

Stakeholder Collaboration

• Serve as the primary technical lead for the workbench workstream, engaging with client data science, architecture, and regulatory stakeholders
• Partner with the imaging platform and data architecture workstreams to resolve cross-cutting dependencies
• Contribute to program-level architecture reviews, technical risk identification, and milestone delivery planning

Required Qualifications

• 10+ years of experience in software or platform architecture, with at least 4 years focused on ML platforms, MLOps, or algorithm development infrastructure
• Strong background in designing MLOps platforms: experiment tracking, model registries, reproducibility pipelines, and deployment workflows
• Hands-on experience with cloud-native ML platform services, including AWS SageMaker, Google Vertex AI, or equivalent managed ML infrastructure
• Experience with statistical computing and data science environments such as Posit (formerly RStudio) Workbench or equivalent platforms used in regulated research settings
• Familiarity with high-performance computing (HPC) environments and GPU cluster architectures for large-scale model training workloads
• Hands-on experience with cloud-native compute environments on AWS, including container orchestration, IAM, storage integration, and GPU resource management
• Experience integrating ML workbenches with upstream data lakes or feature stores, including versioned data access patterns
• Demonstrated ability to deliver within regulated environments, with familiarity in GxP, 21 CFR Part 11, or IEC 62304 requirements
• Strong communication and stakeholder management skills, with experience translating data science and ML requirements into platform architecture decisions

Nice to Have

• Hands-on experience with NVIDIA AI Workbench, NVIDIA MONAI, NVIDIA Clara, or equivalent medical imaging ML frameworks
• Background in medical imaging algorithm development workflows, including segmentation, classification, or detection model pipelines
• Familiarity with FAIR data principles and reproducible research standards in a life sciences context
• Experience with experiment tracking tools such as Weights and Biases (W&B) or MLflow
• Prior experience in a fixed-fee, milestone-based delivery engagement within a large regulated enterprise environment