Why this work matters
Fluent AI output can still be inaccurate, incomplete, poorly targeted, or unsafe. Teams need practical standards that turn subjective reactions into repeatable quality decisions.
02
Applied AI · Quality · Human judgment
AI Training & Evaluation
Structured methods for evaluating educational and marketing AI output, diagnosing weaknesses, improving prompts, and training people to use AI responsibly.
What I contributed
Designed evaluation prompts, weighted rubrics, scoring dimensions, revision logic, and human-in-the-loop review workflows informed by education and marketing expertise.
What it demonstrates
Connects AI capability with domain judgment so organizations can improve accuracy, audience fit, instructional value, brand alignment, usability, and risk awareness.
Selected deliverables
Work products
- 01Educational-content evaluation and improved response
- 02Marketing-campaign evaluation and revised campaign concept
- 03Prompt-engineering comparison across three iterations
- 04Weighted rubric covering quality, audience, ethics, and usability
- 05AI literacy and evaluator-training framework
Evidence
What supports the work
- Prompt, baseline output, scoring rubric, diagnosis, and revision
- Requirements translated into measurable evaluation dimensions
- Original-versus-improved response comparison
- Human approval and documented rationale retained throughout
Evidence status
Original portfolio demonstrations showing evaluation method and decision quality—not benchmark claims about model performance.
Methods & capabilities
Skills used in this work
LLM evaluationPrompt engineeringRubric designAI trainingHuman-in-the-loop review