MLOps components

Deployment e ciclo di vita in MLOps

Nemanja Radojkovic

Senior Machine Learning Engineer

Concepts

GENERIC (Software Development)

Workflows
Pipelines
Artifacts

ML-SPECIFIC

Model registry
Feature store
Metadata store

devops only

devops with mlops

workflow generic

workflow execution

manual execution

automatic execution

semi automatic

before scripting

scripting steps

pipeline definition

artifacts

¹ https://en.wikipedia.org/wiki/Artifact_%28software_development%29

artifacts - definition

¹ https://en.wikipedia.org/wiki/Artifact_%28software_development%29

life start

artifacts to deploy

build it first

one for the model

one for the app

App build pipeline

ML build code only

ML build - empty

raw data

processed data

feature store

pipeline execution

pipe execution 2

trained model

metadata build

model registry

metadata store

deployment pipelines

start from build

deploy to server

monitoring

high level MLOps architecture

Let's practice!

Deployment e ciclo di vita in MLOps

Preparing Video For Download...