AI-native workflow in EDC

EDC uses AI to accelerate technical execution while requiring reviewable, reproducible artifacts.

Where AI is used

• Scoping: framing the project, checking feasibility, defining success criteria
• Data: discovering and accessing relevant cloud-hosted datasets
• Coding: implementing and debugging notebook code, refactoring, documenting
• Interpretation: testing alternative explanations and checking potential failure modes

Required artifacts

• A version-controlled repository (Git/GitHub)
• A reproducible notebook that runs end-to-end
• A short prompt/decision log covering major AI-assisted steps
• Validation checks appropriate to the analysis (baselines, sanity checks, unit checks)

Guardrails

• AI suggestions must be validated, and results require explicit checks
• Students must be able to explain code and reasoning in their own words
• Proper attribution for external code and sources