Machine Learning Engineer

MyOme•Menlo Park, CA

20d•$170,000 - $200,000•Hybrid

About The Position

MyOme is a CLIA-certified clinical whole-genome sequencing company that combines clinical-grade sequencing, AI methods for genome interpretation, and digital tools for clinicians and patients. It is building AI risk models that combine clinical data with genomics to predict disease risk before onset. The Machine Learning Engineer will develop and improve deep-learning risk models and take them from training through deployment as reliable, monitored, clinically reportable services. The role includes applying modern AI, including LLM and agentic methods, where they provide measurable value. The engineer will work with model developers and clinical, regulatory, product, and engineering teams.

Requirements

MS or PhD in computer science, statistics, biomedical informatics, bioengineering, or a related field (PhD preferred), or equivalent experience.
At least 5 years developing machine-learning and deep-learning models, including at least 3 years with models in production.
Strong ML fundamentals, including supervised learning, survival analysis (Cox and deep survival models), calibration, class imbalance, leakage and cohort shift, and evaluation metrics and methods such as AUROC, C-index, calibration curves, and decision curves.
Strong advanced deep-learning model development background, including designing, training, and debugging modern PyTorch architectures such as transformers, sequence and temporal models, multimodal fusion, and deep survival models, with a track record of models that outperform strong baselines.
Rigorous experimental practice, including ablations, hyperparameter search, regularization and overfitting control, reproducibility, and honest comparison against baselines.
Production ML engineering experience with packaging, APIs, containers, CI/CD, testing, model versioning, and monitoring.
Experience building and evaluating LLM-based systems, including prompting, tool use and agents, retrieval, and structured extraction, with quantitative evaluation.
Strong Python and SQL skills, comfort with large tabular datasets, and experience in cloud environments such as GCP or AWS.
Clear communication skills and ability to explain model behavior to clinicians and product partners.

Nice To Haves

Experience with EHR data, including OMOP, ICD, LOINC, RxNorm, and temporal features; experience with All of Us, UK Biobank, or health-system data is particularly relevant.
Familiarity with polygenic risk scores or statistical genetics.
Experience in regulated environments, including CLIA, CAP, HIPAA, FDA SaMD, or design controls.
Experience with MLOps tooling such as MLflow or similar, Airflow or Prefect, and Kubernetes.
Publications or shipped products in clinical ML.

Responsibilities

Develop and improve deep-learning risk models, including architecture design, multimodal EHR and genomics fusion, temporal and sequence modeling of patient histories, deep survival models, representation learning, ablation studies, and benchmarking.
Productionize disease risk models as versioned bundles containing model weights, survival model coefficients, calibration, and PRS scaling, with reproducible training and a model registry.
Build and maintain inference services using Python, FastAPI, and Docker, and integrate them with MyOme's reporting and data platforms.
Implement calibration, recalibration for new populations, and score reporting, including absolute risk over defined horizons, percentiles, and risk tiers.
Establish deployment monitoring for calibration, discrimination, input drift, and coverage, with defined retraining triggers.
Run and automate external validation on partner datasets in secure environments.
Coordinate model evaluation at clinical trial sites with the hiring manager, including site data intake, harmonization, evaluation runs, and results reporting.
Apply LLM and agentic methods where they add measurable value, such as structured extraction from clinical notes, automated evaluation, and guardrailed explanation of results.
Produce validation reports, model cards, and change-control documentation to support CLIA and regulatory review.