Responsibilities
Turn model releases into a pipeline: every new version is evaluated automatically, gated against the one in production, approved, and rolled out with a fast rollback.
Build drift and performance monitoring for our production models, and the retraining that follows from it.
Own the model lifecycle end to end: training pipelines, versioning and the model registry, automated evaluation, promotion and rollback, deployment, monitoring, and retraining.
Build CI/CD for ML: automated testing and validation of every model version, release gates with clear criteria, and an audit trail of who shipped what and why.
Serve models in real time through our Python model-serving service and Snowflake Container Services, with a focus on low latency, reliability, and graceful degradation. Run batch inference in the warehouse with clear standards for latency, correctness, and cost.
Build and operate ML pipelines in Airflow and Snowflake, from feature preparation through training and batch scoring, with reproducible training data.
Monitor drift, data quality, latency, and model performance, and automate retraining that goes through the same release gate.
Partner with data scientists, product engineers, and data engineers. Define clear interfaces between modeling, serving, and the services that consume models, and push back when a request belongs elsewhere.
Handle members' health data responsibly: access control, pseudonymization, and nothing sensitive in logs.
Write high-quality, well-tested production code, and own its observability, debugging, and reliability.
Requirements
5+ years of software engineering, including 3+ years of hands-on MLOps owning production ML systems on AWS with a data or ML platform (e.g., Snowflake ML, Databricks, SageMaker, Vertex AI).
Hands-on model management: versioning, registries, automated evaluation, and promotion and rollback of models in production.
Experience building CI/CD and pipelines for ML: automated training, validation, and deployment (e.g., Jenkins, GitHub Actions), with orchestration in Airflow or similar.
Strong Python experience, including building and operating Python services (APIs or gateways) in production.
Experience serving models in real time: low-latency inference, autoscaling, and reliable serving architectures.
Strong SQL and data warehouse experience (Snowflake preferred).
Experience monitoring models in production (drift, data quality, and performance) and acting on what you find.
Experience owning a solution end to end, from pipeline and model integration through deployment and production debugging.
Proficiency with AI coding assistants as a core part of your workflow, and the judgment to review and verify what they produce.
Nice-to-Haves
Snowflake ML: Model Registry, Snowpark Container Services, ML Observability, or Notebooks and ML Jobs.
Shadow and canary releases for models.
Feature stores and ML lifecycle tooling (e.g., Feast, MLflow, Weights & Biases).
Model development: feature engineering, training, and evaluation with scikit-learn, XGBoost, or LightGBM.
Kubernetes and infrastructure as code (Terraform).
Event-driven systems (Kafka).
Integrating LLMs into production systems.
Observability stacks (Prometheus, Grafana, Datadog).
Experience with health or other regulated data (HIPAA).
Hello Heart has a positive, diverse, and supportive culture - we look for people who are collaborative, creative, and courageous. Oh, and if you want to see some recent evidence of the fun things we do at Hello Heart, check out our Instagram page.
