Pictor Labs is revolutionizing digital pathology adoption worldwide through cutting-edge AI-powered technology. Their DeepStain™ and ReStain™ technologies enable unlimited virtual staining from a single tissue sample, eliminating bottlenecks and limitations of traditional chemical staining processes. This innovation supports the critical evolution from research applications to clinical deployment, empowering laboratories to advance their digital pathology capabilities while reducing chemical waste, improving operational efficiency, and expanding diagnostic possibilities. The role is to own the representation-learning and generative modeling stack that powers Pictor’s virtual staining, building generalizable, production-ready models and evaluations that stand up in clinical workflows.
Requirements
- Representation learning: strong grasp of modern self-supervised methods (contrastive, masked, teacher-student).
- Generative modeling: practical experience with image-to-image translation and controllable generation; comfort reasoning about fidelity, consistency, and robustness.
- Scaling & efficiency: distributed training, optimization for training/inference, and high-throughput data pipelines for very large image dataset.
- Evaluation science: design of rigorous metrics and studies. Proficiency with calibration, uncertainty, and error analysis.
- Data-centric ML: dataset curation, augmentation strategy, artifact handling, label noise management, and reproducibility.
- MLOps fundamentals: reproducible experiments, versioned datasets/models, automated benchmarks, monitored inference, and audit-ready documentation.
- Software engineering: production-quality Python and deep-learning stack fluency; testing, code reviews, etc.
Responsibilities
- Lead pretraining of foundation representations for high-dimensional images; design SSL objectives, curricula, and evaluations.
- Advance image-to-image translation for virtual staining; incorporate uncertainty and calibration.
- Build similarity/retrieval capabilities over learned embeddings for discovery, triage, search, QA.
- Design robust QC and acceptance criteria; detect/mitigate batch effects and domain shift.
- Productionize models with reliable data/experiment tracking, evaluation automation, and monitored serving.
- Partner tightly with pathologists, product, and engineering to turn research into repeatable workflows.
Other
- Opportunity to work with a talented and passionate team at the cutting edge of pathology AI
- Pictor Labs is an equal-opportunity employer.
- We celebrate diversity and are committed to creating an inclusive environment for all employees.
