Nguyen Hanh Phuc

Evaluated Vietnamese AI voice outputs for Outlier’s Live Speech-to-Speech project by ranking model responses using rubric-based criteria (utility, audio quality, conversational dynamics, and naturalness). Also assessed pairwise TTS outputs on Outlier’s Omni TTS Elo project with focus on persona adherence vs. naturalness, and wrote clear, evidence-based English rationales referencing specific linguistic factors such as pronunciation, intonation, and fluency.

Evaluated AI voice outputs and conversation performance for Vietnamese speech-to-speech and speech-related tasks. Planned multi-turn scenario tests for Outlier, classified failures using a defined error taxonomy (e.g., hallucination, failed correction, refusal, persona drift, looping), and wrote evidence-based rationales for reviewer verification. Previously stress-tested hypotheses and built scoring frameworks for expert interviews across 4 industries; communicated findings to senior client leaders and non-technical teams. Assessed pairwise TTS outputs and provided clear, well-justified English evaluation rationales with attention to subtle fluency/pronunciation issues, following structured rubric guidelines.

Researched the U.S. print-on-demand (POD) market using tools such as Merch Intel, Amazon Quick View, and Google Trends. Analyzed product rankings and competitor listings to extract insights on design trends, pricing, content strategy, and customer preferences. Generated 500 optimized listings per month and achieved 300+ sales per month, including a standout product reaching 174 sales and two 5-star reviews within the first 25 days after launch. Integrated data and visual outputs into proposal/deck/report workflows to speed turnaround time.