RAISEF
Case Study 1: AI-Driven Healthcare Diagnostics
Development
Build with traceability, secure defaults, and collectable evidence
Curate data, control features, and version code, configs, and datasets so interpretability, privacy, and bias mitigation stay practical.
In Development, the design becomes working systems that are easy to inspect and audit. Datasets are curated with provenance, quality checks, and coverage notes. Features and prompts are controlled and documented so unwanted proxies are less likely. Code, configurations, and data are versioned together, with reproducible runs and logs that tie results to people and commits. Access follows least privilege, secrets are managed safely, and automated checks enforce style, tests, and basic security. Instrumentation captures the indicators chosen earlier, and model cards or equivalent records are drafted as work progresses. Clear assumptions and risk notes prepare the ground for rigorous testing, evaluation, and human oversight in the next stage.
