## MLOps --- ## MLOps for UMC - Present: Dragon Chen - Date: 10/16 --- ## What is MLOps {%youtube _jnhXzY1HCw %} ---- <!-- launching is hard -->  ---- <!-- researcher and engineer -->  ---- <!-- devops -->  ---- <!-- mlops -->  ---- <!-- lifecycle -->  ---- ## For me - kind of soft skill of ml - Management of Machine Learning System - Build smart model, build model smart --- ### Key to ml is the key to engineer simple code, make readable, comment, test, fix ever-present sign mistake, peer review, version controal, track performance online and launch iterately ----  --- ## Where is Solution - Gitlab repo + CI/CD (Code version and test) - Kubernetes and Kubeflow, argo (Flow) - Mlflow (Model Versioning) - DVC (Data Version Control) - R shiny, Streamlit (development + serving) ---- ### Gitlab + CI/CD  Test, Automtion and Tracking **Loop of your "code file" or "code run"** ---- #### DVC  Support many kinds of Storage ---- ### MLFlow --- ### Demo