01AI EVALUATION · TASK DESIGN · QUALITY ASSURANCE
Real-world workflows. Ambiguous requests. Clear evaluation, backed by evidence.
Real problem, unclear inputs.
→Find and verify the right sources.
→Resolve gates, trade-offs and dependencies.
→Define a clear, reviewable success state.
Evaluation tasks
Contributor submissions reviewed
Financial-services scenarios
Quality rating
Realistic scenarios. Explicit constraints. Verifiable expected outcomes.
View all workHow I turn evaluation criteria into consistent, defensible judgments.
Understand the task, context, and what a successful response or run should look like.
→Translate guidelines, policies, and evidence into clear grading criteria.
→Assess the model’s behavior against the criteria, constraints, and expected outcome.
→Resolve ambiguity and reach a consistent, defensible judgment.
Hands-on experience across operations, finance, procurement and cross-functional delivery helps me distinguish a plausible answer from one that would actually hold up in a live workflow.