Assessment Guidelines

Overview

Your final assessment is based on the Air Quality project you build throughout Sessions 2 and 3, and on an individual supervised exam. Together, they show your ability to move from exploratory data science to structured, maintainable and well-documented machine learning work, and to explain what you have learned to someone else.

Core competencies

Your work will be assessed based on three fundamental competencies that align with the official program objectives:

• [BC04] Software Engineering Best Practices: Apply professional development principles to data science workflows, ensuring code quality, maintainability, and effective collaboration.
• [BC07] End-to-End ML Pipeline Design: Design and implement complete, production-ready machine learning pipelines using modern MLOps practices and industry-standard tools.
• [BC02] Data-Driven Communication: Transform technical analysis into clear, actionable insights suitable for both technical and business audiences through effective visualization and statistical analysis.

Each competency is broken down into critical learnings with three levels of mastery (Novice, Intermediate, Competent); see the Competency Levels page to find out what is expected of you at each level, the Competent level being the target to reach by the end of Session 3.

Grade Determination

All assessment is individual. Your final grade is decided in two steps: the number of Session 3 modules you complete sets the highest grade you can reach, and you obtain that grade only if all three competencies are validated. If any competency is not validated, your grade is F and the course is not passed.

Grade ceiling

A module counts as completed only if its reflection file contains a substantial answer (see Module Completion).

Competency validation

BC04 is validated automatically from your Git repository: the unit tests of every module you completed must pass, your commit messages must follow the Conventional Commits format, and your history must show regular, incremental work.

BC07 and BC02 are validated during the supervised exam described below.

The supervised exam

The exam is based on a problem drawn at random from the Session 3 modules you have completed. The more modules you complete, the higher your possible grade, but the less you know in advance which one you will get. The exam has two parts:

1. Notebook (2h30). Using the provided template, write a notebook that explains the problem to someone who has not followed the module. BC07 checks that you select the code and data that make the problem visible and show how to solve it. BC02 checks that your explanation is simple, clear, and actually teaches the reader something.
2. Peer review (1h). Review four notebooks written by other students with the provided grid, and justify each review.

A competency is validated when all four reviewers agree that it is. In every other case, a teacher reviews the notebook and decides; teachers may also check notebooks that were validated unanimously. If your own reviews clearly differ from the teacher’s decision, showing leniency toward poor work or a clear misunderstanding, your grade is capped one level below your grade ceiling.

Module Completion

Each module’s written reflection is its deliverable of record. In your air_quality project:

• Session 2 (mandatory, in order): reflection/session_2/module_1.md, module_2.md, module_3.md
• Session 3 (optional, any order): reflection/session_3/module_1.md through module_9.md, one per module you completed

A missing reflection file, or one without a substantial answer, means that module is not counted as completed, regardless of what the corresponding notebook or source code shows. Teachers read these reflections to check that they make sense.