Engineering Production-Ready ML/AI Systems

Course manual 2025/2026

Course content

The field of ML/AI made significant impact on the organizations, solving complex business problems. However, going beyond proof-of-concepts and bringing ML/AI systems in the hand of users remains a challenge. Engineering production-ready ML/AI systems is crucial in the success of ML/AI initiatives but often challenging as it requires a scarce set of skills from multiple domains.

In this course, we explore (1) how production-ready ML/AI systems are being built, (2) what are the organizational and technical challenges of realizing ML/AI projects, and (3) how ML/AI discipline can employ modern software engineering principles to facilitate successful ML/AI initiatives. This course has both practical and conceptual elements.

Study materials

Literature

• "Machine Learning in Production: From Models to Products" by Christian Kästner

• "Outcomes Over Output: Why customer behavior is the key metric for business success" by Joshua Seiden

• "Clean Code" by Robert C. Martin

• "Designing Data-Intensive Applications" by M. Kleppmann

• "Designing Machine Learning Systems" by Chip Huygen

Syllabus

• Syllabus in Canvas

Software

• Python IDE of your choice - Pycharm recommended. 

• Apache Spark

• Docker

• Apache Airflow

• Prometheus

Other

• Various articles, posts, and guidelines that can be found in Canvas.

Objectives

• Students will be able to explain and analyze ML/AI related challenges in corporate environments.
• Students will be able to analyze ML/AI related code for simplicity, clarify, and readability.
• Students will be able to implement automated testing at unit and system-interaction level.
• Students will be able to understand distributed data processing concepts.
• Students will be able to create, inspect, and work with containerized ML/AI applications.
• Students will be able to differentiate different ML/AI deployment strategies. Students will be able to create batch ML/AI jobs, and HTTP API's for ML/AI applications.
• Students will be able to design and implement telemetry elements for ML/AI systems.

Teaching methods

• Lecture
• Laptop seminar
• Working independently on e.g. a project or thesis

Lecture: teaching conceptual elements and demonstrating example implementations.

Laptop seminar: Working on complex problems with the guidance of TA's.

Project: Implementing an end-to-end solution independently.

Learning activities

Attendance

This programme does not have requirements concerning attendance (OER part B).

Assessment

• Assignments are 60% of the course's grade.
• Course project is 40% of the course's grade.
• You must have at least 55% (5.5) in course project to pass.
• Grading for each assignment and course project is described in the Assignments in Canvas.
• Failing to meet the deadline for a particular assignment or project will grant 0 points. There will be no extensions.

Inspection of assessed work

TA's will make the grading available on Canvas. Students will have 1 week period for that assignment where they can ask questions regarding the assessed work.

Assignments

See Canvas for detailed and up to date information.

• Students expected to work on all assignments and the course project individually. You must not share any work with other students.
• Readings are optional, but highly recommended. They will help you during your assignments and the course project. Readings will not count towards the course grade.
• You are expected NOT to use any (Gen)AI tool.

Fraud and plagiarism

The 'Regulations governing fraud and plagiarism for UvA students' applies to this course. This will be monitored carefully. Upon suspicion of fraud or plagiarism the Examinations Board of the programme will be informed. For the 'Regulations governing fraud and plagiarism for UvA students' see: www.student.uva.nl

Course structure

Contact information

Coordinator

• K.T. Yesilbek