1. Domain knowledge
  2. Requirements
  3. Data / Rules
  4. AI System
  5. Evaluation
  6. Feedback
  7. Improvement

Engineering approach

That means thinking about retrieval, evaluation, state, failure handling, data quality and what the system should do when evidence is insufficient.

Where the work points

This has pulled my work toward production AI, evaluation, retrieval and the data systems underneath them. A system is reliable when failure is visible, measurable and recoverable.

Contact