Portfolio: AI Engineering Assessment Concept

Scenario: "AI Research Lab Anomaly" - Candidates investigate mysterious model behavior at a fictional AI research facility.

Challenge Design

• Debug a "haunted" neural network exhibiting unexplained behaviors

• Trace data lineage through corrupted training logs

• Collaborate with "AI researchers" (other candidates) via simulated Slack

• Present findings in a mock incident review

What We Measure

• Systematic debugging approach vs. random experimentation

• Ability to form and test hypotheses

• Communication clarity under pressure

• Willingness to acknowledge uncertainty

• Collaboration vs. competition dynamics

Pilot Results

Tested with 12 senior ML engineers. The narrative framework increased completion rates from 41% (traditional take-home) to 92% (ARG format). Candidates reported the experience as "actually fun" and "more relevant than coding challenges."