Course
Digicomp Code H36447
MLOps in Practice: Deployment and Integration of Machine Learning Models («H36447»)
Course facts
- Gaining in-depth knowledge of MLOps (machine learning operations) concepts and methods
- Learning about basic concepts and tools and gaining practical experience working with the most important tools (DVC, Dagster, MLflow, FastAPI, ONNX, and many more)
- Acquiring valuable skills for designing, planning, implementing, and maintaining scalable data and machine learning pipelines
- Gaining in-depth theoretical knowledge of the operationalization of machine learning models as well as practical experience in applying the methods and tools
- Developing, customizing, monitoring, and productively deploying your own machine learning pipelines
1 MLOps – what it is and why you can't do without it
- When machine learning projects get serious
- Domain knowledge and challenges
- The MLOps cycle at a glance
- MLOps is more than DevOps
- The MLOps maturity levels
2 Data versioning and experiment tracking
- Basics and advantages of code and data versioning
- Introduction to DVC
- Exercise: Data versioning with DVC
- Exercise: Experiment tracking with DVC
3 Data pipeline orchestration
- Basics and advantages of data pipelines
- Introduction to Dagster
- Exercise: Asset jobs with Dagster
- Exercise: Op jobs with Dagster
4 Experiment tracking
- Parameters, metrics, and artifacts
- Basics and advantages of experiment tracking
- Experiment tracking with MLflow
- Exercise: Experiment tracking with MLflow
- Exercise: Model management with MLflow
5 CI/CD for machine learning
- Introduction to CI/CD, differentiation from CI/CD for code
- What can we test?
- Variants of CI/CD for ML products
- Showcase: Github Actions and CML
6 Deployment and serving
- Basics of machine learning deployment
- Differentiation between batch inference and live inference
- Data preprocessing in deployment
- Introduction to Open Neural Network Exchange (ONNX)
- Exercise: FastAPI and ONNX
7 Monitoring
- Monitoring ML models
- Data, metrics, KPIs
- Application metrics
- Showcase: Monitoring with evidently.ai
8 MLOps in the cloud
- When are cloud solutions recommended?
- Classification of Amazon Sagemaker, Azure ML Studio, and Google Vertex AI
- Showcase: Model training with Azure ML Studio
9 Machine learning platforms
- How and when do I scale the development of my ML teams?
- What is a feature store?
10 Excursus: LLMOps
- What distinguishes LLMOps from MLOps?
- Showcase: companyGPT
This online seminar will be held in a group of no more than 12 participants using Zoom video conferencing software.
The instructors will be on hand to assist you with the practical exercises—either in the virtual classroom or individually in breakout sessions.
Once you have registered, you will find all the information, downloads, and extra services relating to this training course in your online learning environment.
This training is aimed at anyone who wants to create, operate, monitor, expand, and fine-tune machine learning models and applications. This course is a valuable building block in the qualification process to become an MLOps expert, machine learning engineer, data scientist, and data engineer.
Basic technical knowledge of machine learning models and algorithms is required. Prior knowledge of mathematics and statistics is helpful but not required.
To ensure that you receive any necessary documents by mail in good time, we recommend booking at least 14 days before the seminar date.