Principal Software Algorithm Engineer - Digital Cytology AI

Hologic•Marlborough, MA

About The Position

Hologic is seeking a Principal Software Algorithm Engineer to design, train, and validate deep learning models for cytology on the Genius™ Digital Diagnostics System. The role owns the cell-level CNN classifier, the nuclear-to-cytoplasmic (NC) ratio segmentation pipeline, and technical evidence supporting FDA and other regulatory submissions. The engineer serves as technical lead for model design and verification under design controls, collaborating with cytopathologists, imaging scientists, and regulatory and quality teams. The position also involves establishing model performance and interpretability practices, mentoring engineers, and setting coding, reproducibility, and MLOps standards. The annualized base salary range is $157,000–$220,000, and the role is bonus eligible. Final compensation depends on relevant experience, skills, knowledge, education, business needs, and market demand.

Requirements

MS or PhD in computer science, biomedical engineering, electrical engineering, or a related field, or a BS with equivalent experience.
10 or more years of software development experience, including 5 or more years building and deploying deep learning models for computer vision.
Demonstrated leadership experience setting strategic direction, influencing senior stakeholders, leading cross-functional initiatives, and driving accountability and execution in a complex, matrixed organization.
Expertise in Python and PyTorch or TensorFlow, including CNN fine-tuning, transfer learning, and segmentation.
Experience with imbalanced, multi-class, and multi-task problems, and metrics such as per-class sensitivity, F1, kappa, and AUROC.
Experience with medical or microscopy imaging, including color handling, stain variation, and high-resolution tiling.
Strong discipline in data splitting, leakage prevention, experiment tracking, and reproducible training.
Clear written communication for technical reports and design documentation.

Nice To Haves

Digital pathology or cytology AI experience, ideally urine cytology.
Experience with FDA 510(k), De Novo, PMA, or EU IVDR.
Familiarity with IEC 62304, ISO 13485, ISO 14971, 21 CFR Part 820, and FDA PCCP guidance.
ICC color management, stain deconvolution, and scanner color calibration experience.
Familiarity with cell instance segmentation tools such as StarDist and Cellpose, and annotation tools such as QuPath, CVAT, and napari.
Model optimization for edge or embedded inference, including quantization, ONNX, and TensorRT.
Experience leading or technically guiding a small team.

Responsibilities

Design, train, and tune multi-class CNNs to classify cells and aggregate predictions into specimen-level diagnostic categories.
Develop multi-task models with shared encoders and feature heads for diagnostic class, cytomorphologic features, and NC ratio category.
Build nucleus and cytoplasm segmentation networks to measure NC ratio and nuclear area, including in overlapping cells in clusters.
Own the data pipeline, including patient-grouped splits, color profile application, class balance, and imaging metadata.
Run disciplined experiments, benchmark lightweight backbones against larger reference models, and define evaluation metrics and reports.
Collaborate with pathologists on labeling protocols, inter-observer agreement, and adjudication processes.
Prepare design documentation, verification and validation protocols, and technical content for regulatory submissions.
Mentor engineers and establish coding, reproducibility, and MLOps standards.
Ensure model performance and interpretability, including Grad-CAM checks and feature outputs aligned with pathologist expectations.
Support regulatory strategy for AI-enabled medical devices under design controls and PCCP guidance.
Lead model design and verification under design controls in collaboration with cytopathologists, imaging scientists, and regulatory and quality teams.