At a glance
Certificate of attendance "MLOps" (2 ECTS)
55h: 16h Lecture, 30h Project work, 9h Self-study
Language of instruction:
Objectives and content
This course is aimed at data scientists, data engineers, product owners, project managers, decision makers, software engineers, researchers, and/or research managers from all fields, who would like to either deploy their own use-case or in any form assist in model development and productization stage of an ML system.
MLOps is defined as a set of practices that aim to deploy and maintain machine learning models in production reliably and efficiently. It is a very broad field that requires an understanding of various aspect of an ML system involving right use-case selection (business + data strategy), data acquisition pipeline development (data engineering), PoC (data scientists/ML researcher), creating a production ready pipeline (end-user based) and maintenance of that system.
Defining MLOps best practices around data, code and model has become a key component in becoming data-driven organizations. This requires coming up with processes to track different stages of model planning and development. In this course we will develop a unique MLOps canvas and use it to define technical and business requirements around MLOps for our use-cases.
There course will be delivered in form of various questions end-users should ask when thinking about MLOps:
- What is MLOps and why is there a need for it?
- What do we want to achieve with MLOps?
- How to define MLOps of an organization?
- How to define a use-case and its data acquisitions strategy?
- What kind of model and testing infrastructure is needed and why?
- What is data drift and how to track it?
The module covers the following topics:
- MLOps Introduction to MLOps Canvas (what and why?) and Use-case planning, including a data acquisition strategy.
- Experimentation to Production ready pipeline
- Production and Maintenance
- MLOps challenge – Who defines the most steps?
There are 4 lessons organized once a week on Tuesday Afternoon. After 3 lessons the students work on their final project for about one week. The final project presentation will be held during the last lesson.
CAS in Digital Life Sciences
This module is part of the CAS in Digital Life Sciences continuing education programme, but can also be attended independently of the CAS. Credit points earned for this module can be credited to the CAS course at a later date, provided the relevant general conditions are fulfilled.
More information here: CAS in Digital Life Sciences
Prior knowledge in machine learning, product development, data engineering, DevOps, and software engineering are useful but not required.
The module will consist of lectures and practical exercises. In addition to lectures, students will be required to interact with each other in groups to define requirements around their use-cases using the proprietary MLOps canvas developed for this training. On the last day students will present their strategy and taken to solve real world problems.
- Exercises during the course: 50%
- Use-case challenge: 50%
Enquiries and contact
Dr. Nitin Kumar
Nitin Kumar has over a decade experience in creating machine learning based business applications in his roles as a data scientist, educator and consultant. He has delivered around 100 Data Science proof-of-concepts and a dozen end-to-end data products. In the last couple of years his focus has been on designing and delivering ML trainings on topics including machine learning for managers, use-case identification and MLOps to push the pace of ML productionisation across several industries including insurance, transportation, manufacturing, and others.
|Start||Application deadline||Registration link|
|on request||on request||Application|