Principal Data Scientist

Responsibilities

• Design, prototype, and rigorously evaluate novel classifier architectures for clinical diagnostics across oncology indications
• Lead exploratory research into new quantification, normalization, and feature engineering methods for high-dimensional glycoproteomic data
• Bring a diverse modeling toolkit — classical statistical methods, tree-based ensembles, deep learning, probabilistic and Bayesian approaches, foundation models, graph neural networks, and generative AI — and choose the right tool for the problem based on evidence rather than habit or hype
• Develop cross-validation, calibration, and uncertainty-quantification strategies that hold up to the realities of small clinical cohorts and high feature counts
• Investigate and mitigate batch, cohort, and site effects so that models generalize from discovery to bridging to locked panels
• Drive cross-indication synthesis — separate shared disease biology from indication-conditioned signal, and from nonspecific inflammatory or acute-phase axes
• Build multimodal models that combine glycan/motif information, proteomic grounding, and clinical covariates rather than relying on protein-quantity signal alone
• Translate emerging techniques from the ML, AI, and computational-biology literature into production-ready methods
• Mentor junior data scientists and raise the methodological bar across the team

Qualifications

• Ph.D. in Statistics, Computer Science, Computational Biology, Bioinformatics, or a related quantitative field, plus 6+ years of experience building predictive models on biological data in industry or academia; alternatively, an MS in a similar field with 8+ years of relevant experience
• Demonstrated track record of methodological innovation — first-author publications, novel methods deployed in production, open-source contributions, or comparable evidence of original work
• Deep proficiency in Python and/or R, including the modern ML stack (scikit-learn, PyTorch or TensorFlow, XGBoost/LightGBM, and similar)
• Methodological breadth across paradigms — comfortable moving between classical statistics, tree-based ML, deep learning, and modern AI (transformers, graph neural networks, foundation models, generative methods) — and the judgment to argue rigorously for one approach over another
• Strong statistical foundation: cross-validation strategy, regularization, calibration, uncertainty quantification, and handling of confounders and class imbalance
• Hands-on experience building and validating classifiers on high-dimensional, low-sample-size biological data (proteomics, glycoproteomics, transcriptomics, or genomics)
• Experience with batch-effect correction and normalization techniques, and a healthy skepticism about how those choices propagate into downstream performance estimates
• Prior experience in the clinical diagnostics industry with a solid understanding of analytical and clinical validation, locking classifiers, and bridging studies
• Excellent written and verbal communication: able to explain novel methods clearly to wet-lab scientists, clinicians, and fellow statisticians alike
• A genuine desire to impact patient lives and contribute to the broader scientific community