Day-to-Day Responsibilities
Build and maintain automated pipelines that load data, train, test and evaluate models, using tools such as Kubeflow Pipelines, Apache Airflow or Prefect.
Set up CI/CD so that model and code changes are tested and deployed with little manual work, using GitHub Actions, GitLab CI/CD, Jenkins or ArgoCD.
Package models as containers with Docker and run them on Kubernetes or cloud services such as AWS, GCP or Azure.
Version code, datasets and models with Git, DVC and MLflow so that any past result can be reproduced and audited.
Monitor live models for latency, throughput, errors and accuracy, and watch for data drift and concept drift using Prometheus, Grafana or the ELK Stack.
Trigger retraining when performance drops, and work with data scientists to move improved models into production safely.
Secure data, model artefacts and infrastructure through access control, encryption and regular audits of third-party dependencies.
