| Dozent/in |
Matthias Egli, MSc; Adrian Willi, MSc |
| Veranstaltungsart |
Vorlesung |
| Code |
HS261022 |
| Semester |
Herbstsemester 2026 |
| Durchführender Fachbereich |
Wirtschaftswissenschaften |
| Studienstufe |
Master |
| Termin/e |
Do, 17.09.2026, 16:15 - 20:00 Uhr, HS 7
Do, 01.10.2026, 16:15 - 19:00 Uhr, HS 7
Do, 15.10.2026, 16:15 - 20:00 Uhr, HS 7
Do, 29.10.2026, 16:15 - 20:00 Uhr, HS 7
Do, 12.11.2026, 16:15 - 20:00 Uhr, HS 7
Do, 26.11.2026, 16:15 - 20:00 Uhr, HS 7 |
| Umfang |
2 Semesterwochenstunden |
| Turnus |
bi-weekly |
| Inhalt |
The field of data science has experienced a renaissance due to innovations in algorithms and the widespread availability of affordable storage and compute capabilities. More recently, the rapid rise of generative AI and its ongoing impact across industries have further reshaped how data-driven software is built and operated. As a consequence, the growing, global stream of data has emerged as a significant economic factor. Nonetheless, many companies struggle to make use of their data. A significant reason for this is a lack of experience in organizing data and software as well as managing a data science team in a collaborative setting. This course sets off, where most data science courses end. It addresses technical and organizational challenges that are typically accompanied by operating data-driven software products in "production". In this context, the course aims to provide solutions for the aforementioned challenges. This includes toolkits and architectures that:
• render the management of data science projects more efficient
• allow for versioning of data, software and runtime environments, in order to ensure reproducibility of data-driven systems
• improve collaboration and knowledge transfer among members of a larger data science team
• facilitate the deployment of data-driven products
In an era where working code is increasingly cheap to generate, the scarce and valuable skills are: reasoning about system design, diagnosing production failures, ensuring reproducibility, and justifying architectural decisions under real constraints. The course emphasizes these skills through hands-on project work, debugging exercises with pre-built systems, and individual oral examinations in which students defend their design choices. |
| Lernziele |
• Understanding of the larger complexity of data-driven software compared to "traditional" software
• A firm grasp of the typical life cycle of machine learning projects in industry
• The ability to reason about and justify tool and architecture choices under given constraints, rather than merely implementing them
• Diagnostic skills for identifying failure modes in ML systems
• An overview of existing toolkits that address the challenges of data-driven products
• Knowledge in a subset of those toolkits that cover different areas, such as:
o code versioning (f.e. Git)
o data versioning (f.e. DVC)
o runtime versioning (f.e. Docker)
o testing frameworks
o ML experiment tracking tools (e.g. Weights & Bias, MLflow)
o production environments for machine learning models
• The ability to maintain ownership and deep understanding of systems built with modern development tools (ersetzt: The students are expected to be able to create a workflow for the development of complex data science products)
• Skills in reviewing and critiquing ML systems built by others |
| Voraussetzungen |
• Experience with Python or R scripts
• Experience in training machine learning models (e.g. linear regression)
• First experiences with the command line (Unix and Windows) |
| Sprache |
Englisch |
| Begrenzung |
max. 25 participants
If the maximum number of participants is reached, students of the MA in Economics and Management will be given priority. In this case, please contact the wf@unilu.ch |
| Anmeldung |
Binding registration takes place via the UniPortal; see notes below in the «Proof of Performance» field.
For course information and materials, registration on the OLAT e-learning platform is required from 31 August – 13 September 2026. The students themselves are responsible for verifying the course’s creditability towards their degree program.
Direct link to the OLAT course: https://lms.uzh.ch/url/RepositoryEntry/17903616219
|
| Leistungsnachweis |
***IMPORTANT*** To receive course credit and earn academic credits, registration via the Uni Portal from 31 August (starting at 9:00 a.m.) – 13 September 2026 is MANDATORY.
Late registrations and withdrawals will not be accepted. Once the registration period has ended, participation in the course is MANDATORY. If the course is not completed without a valid reason and without proper withdrawal (including supporting documentation; see the Exam Guidelines), the course will be considered failed (grade 1).
Assessment consists of two parts: project work (team-based ML system with decision log and reproducibility documentation), and an individual oral examination in which students defend their system's design and demonstrate component-level understanding. |
| Abschlussform / Credits |
Written project report / Oral examination / 6 Credits
|
| Hinweise |
Project work with oral defense. |
| Hörer-/innen |
Nach Vereinbarung |
| Kontakt |
matthias.egli@doz.unilu.ch / adrian.willi@doz.unilu.ch |
| Anzahl Anmeldungen |
0 von maximal 25 |
| Literatur |
The Hundred-Page Machine Learning Book (Andriy Burkov)
Designing Machine Learning Systems (Chip Huyen) |