Practical Skills, Project Work and Employability / Entrepreneurship
Domain 10 - Practical Skills, Project Work and Employability / Entrepreneurship. This domain tests laboratory discipline, honest project reporting, and realistic career and business framing rather than ambition statements.
Practical file rules
- Every practical needs aim, tools, theory, procedure, observations or code, result and conclusion. Code must be reproducible: fixed random seed, named dataset version, noted library versions.
- Record the data source and consent. Personal data must never be uploaded to a public service without permission.
Project structure
- Problem statement → stakeholder and scope → data → method → baseline → results with metrics → limitations and ethical risks → future work → demonstration.
- A strong problem statement is specific and measurable, for example flag late homework by Friday with recall at least 0.7, rather than the vague use AI in school.
Employability and entrepreneurship
- Entry roles: data annotator, data analyst intern, junior Python or ML developer, AI testing and support. Skills: Python, SQL, spreadsheets, statistics, visualisation, Git, clear writing and domain knowledge.
- A portfolio beats certificates alone: two or three documented GitHub projects with README, data card and results.
- Entrepreneurship: identify a real pain point, state the value proposition, build a minimum viable product, estimate cost and pricing, check privacy compliance and IP, then pilot. The Business Model Canvas (customer segments, value proposition, channels, revenue, costs) is the standard framing.
Worked example
Project: predict late homework submission. Reframe as binary classification with labels from the school LMS. Features: past submissions, attendance, subject. Baseline: majority class. Model: logistic regression, evaluated on recall for the late class, because missing a genuinely late student costs more than a false alarm. Ethics: inform students, allow opt-out, never publish individual predictions. Report limitations: one term of data and a single school.
Traps that cost marks
- Vague aims that cannot be evaluated.
- Reporting only accuracy on an imbalanced target.
- Screenshots with no code or reproduction steps.
- Claiming a model proves causation, or using a dataset without checking its licence.