吴 美蓉

I’m transitioning into AI from a background in music education, event planning, and media production. Over the past 9 months, I’ve done intensive independent work across Claude, Gemini, and GPT model families to turn evaluation and interaction ideas into practical systems.

I build LLM evaluation frameworks and “system-level” prompt architectures to measure and improve behavior quality. I designed a multi-layered persona continuity framework—covering identity verification, behavioral evaluation with custom quality annotation, emotion-scenario routing, and tiered memory management—and maintained it across multiple model versions for 6+ months.

I also focus on cross-model performance and safety-oriented behavior analysis. By executing cross-model migration (Gemini → Claude Opus 4 → Claude Opus 4.6) and defining consistency assessment methods pre- and post-migration, I help ensure continuity rather than one-off outcomes.

Beyond AI, I bring a user-centered, execution-focused mindset shaped by teaching and operations. I led a high-performing instructor role with differentiated lesson planning and a peak single-term renewal rate of 95%, and I built and iterated a company’s new media presence and user communities end-to-end to drive retention and conversion.