Loading page
BizPlans
How to Build a Data Annotation Portfolio
Kenya-first insights, practical and grounded.
Back to insights
Published 03/08/2026 - 1 min read
How to Build a Data Annotation Portfolio
Many platforms rely more on assessments than portfolios, but a small body of original work sharpens your skills and supports direct applications. It must never contain client tasks, leaked rubrics or confidential screenshots.
Create four original samples
Classification set: write 20 original short texts, define 3–5 labels, annotate them and explain difficult boundaries.
Side-by-side AI evaluation: use two outputs you created specifically for practice or material you have permission to use. Score factuality, relevance, clarity, instruction-following and safety. State why one wins.
Fact-check sample: choose a public claim, find primary sources, separate verified facts from inference and record publication dates.
Domain sample: if you are a coder, review a small open-source snippet under its licence and write tests. If you are a language specialist, annotate original Swahili or Somali examples for tone, intent or ambiguity.
Publish a short methodology, label definitions, examples, quality-control pass and limitations. Remove personal data. Keep the whole portfolio to a few polished pages or a simple public repository.
Respect platform rules
Do not copy assessment questions or reuse work from one annotation platform on another. Handshake's policy, for example, says project contributions must be original and bars prompts taken from textbooks, learning sites or other annotation work; it also prohibits external AI tools unless authorized. Always follow the stricter project-specific rule.
For original exercises and portfolio templates, see the AI Training & Data Annotation Work Guide for Kenya.
Sources
Next step
If you are ready to turn the idea into an execution plan, browse the downloadable guides or generate a custom plan for your business model.
Use the full model