Senior ML Engineer – Healthcare

The mission Help us build the labeling and outcomes backbone for a healthcare analytics platform. You’ll turn messy operational/clinical signals into auditable, programmatically generated training labels—so downstream models can be trained and evaluated safely, with governance and privacy in mind. What you’ll do • Own the outcomes registry: design schemas and pipelines that store labeling-function (LF) votes, model/posterior scores, thresholds, and reviewer adjudications with as-of time safety and clear lineage. • Stand up weak supervision: author 10–15 LFs, train a label model (e.g., Snorkel/“data programming”), and ship coverage/precision dashboards for stakeholders. • Quality & governance by design: add validation tests (schema/range/leakage), version everything, and emit immutable audit events for reproducibility. • Gold set workflow (small, surgical): operationalize a lightweight review process (two reviewers + adjudicator) to calibrate and verify weak-label quality. • Partner with modeling: produce tidy, well-documented outputs (LF votes, posteriors) that modelers can plug into training & evaluation jobs. Qualifications (must-have) • 4+ years in ML/Data Science or MLE building production data/label pipelines. • Hands-on Snorkel (or equivalent weak supervision frameworks) and label noise handling. • Strong Python (Pandas/NumPy), PyTorch, and SQL; comfortable with MLflow & Great Expectations. • Experience designing governance/audit-friendly data models (versioning, reproducibility, immutability). • Comfort translating clinical heuristics or policies into deterministic rules. • Deep understanding of as-of time windows and leakage prevention. Nice-to-have • Healthcare data (EHR/claims/LTC) and HIPAA familiarity. • Federated learning exposure (Flower/FedProx) and/or representation learning on tabular sequences. • Azure (AKS, Key Vault), Docker, GitHub Actions/Azure Pipelines.