Machine Learning Engineer ll

📋 Description

• • Own the end-to-end lifecycle of production-grade ML models that surface clinically-relevant insights from millions of structured and unstructured healthcare records, directly influencing prior-authorization decisions and patient care pathways.
• • Architect, train, and fine-tune transformer-based NLP models to extract entities (diagnoses, procedures, medications, lab values) from physician notes, radiology reports, and EHR free-text with >95 % precision/recall on HIPAA-compliant data.
• • Design scalable feature pipelines that blend claims, HL7 FHIR feeds, and narrative text into real-time feature stores (Redis, DynamoDB) consumed by downstream ranking and risk-prediction services.
• • Implement robust model-serving infrastructure on Kubernetes (EKS) with canary releases, A/B gates, and automated rollback policies to guarantee 99.9 % uptime for API endpoints processing 5 k+ requests/minute.
• • Build continuous-training loops using Airflow and SageMaker Pipelines that retrain models weekly on fresh labeled data, monitor drift via statistical and embedding-based detectors, and trigger human-in-the-loop review when thresholds are breached.
• • Collaborate cross-functionally with clinical informaticists, product managers, and regulatory teams to translate complex medical policy rules into probabilistic constraints and fairness objectives embedded in model loss functions.
• • Create interpretability dashboards (SHAP, LIME, attention heat-maps) that allow medical directors to audit model reasoning and satisfy NCQA, CMS, and state-level compliance audits without exposing PHI.
• • Optimize GPU utilization and inference latency by experimenting with mixed-precision training, ONNX export, TensorRT compilation, and dynamic batching, cutting cloud spend by 30 % while maintaining sub-100 ms P95 latency.
• • Establish data-quality SLAs and anomaly-detection monitors that alert on schema drift, missing concepts, or unexpected demographic skews, ensuring equitable performance across age, race, and payer groups.
• • Mentor junior ML engineers through code reviews, design docs, and weekly learning sessions on topics like federated learning, differential privacy, and causal inference in healthcare.
• • Contribute to the open-source community by publishing anonymized datasets, releasing model cards, and presenting at conferences such as MLHC, NeurIPS Health Workshop, or ACM CHIL.
• • Drive experimentation culture by maintaining a centralized MLflow registry with reproducible notebooks, hyper-parameter search tracking, and automated model performance reports that feed directly into sprint planning.
• • Partner with security engineers to implement PHI tokenization, field-level encryption, and SOC-2-compliant logging so every model artifact and dataset meets HIPAA, HITRUST, and GDPR standards.
• • Influence product roadmap by translating clinical KPIs (denial overturn rate, care-gap closure, prior-auth turnaround time) into measurable ML objectives and quarterly OKRs.
• • Champion ethical AI practices by embedding bias-mitigation techniques (re-weighting, adversarial debiasing) and fairness dashboards into every release cycle, ensuring equitable outcomes for underserved populations.