懐偉銘

Key Skills

Software

Don't disclose

Top Subject Matter

Livestream commerce and cross-border product selling content

Cross-border e-commerce product information and localized customer messaging

LLM response evaluation and prompt/script refinement for livestream commerce

Top Data Types

Text

Image

Top Task Types

RLHF

Freelancer Overview

Livestream Host — AI training fit through response fit evaluation and prompt/script iteration (Zhejiang Shiyang Knitwear. Brings 6+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Don't disclose, Codex, and Coze. Education includes Bachelor of Business Administration, Senshu University (2023). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Translation.

Labeling Experience

Operations Lead / AI Trainer-aligned quality reviewer — output evaluation and feedback loop creation (Wuyou Media, Hangzhou)

TextRLHF

Built a practical quality-standard workflow for evaluating outputs, diagnosing failure points, and iterating language to improve user response and business outcomes. Used an evaluation mindset to compare generated responses across quality dimensions such as completeness, tone, risk, and business fit. Converted livestream scripts, customer feedback, and creator inputs into reusable rules, examples, and improvement notes for continuous refinement loops.• Defined quality standards and reviewed outputs to identify failure modes.• Iterated messaging based on engagement and conversion data from multiple livestream rooms.• Coordinated creators and brands while aligning scripts and interaction styles to target users.• Documented improvement notes from real sessions to support ongoing model/response improvement.

2024 - 2026

Livestream Host — AI training fit through response fit evaluation and prompt/script iteration (Zhejiang Shiyang Knitwear)

Don't discloseText

Developed and applied evaluation criteria for livestream responses by judging whether answers matched user context, including intent and relevance. Reviewed audience, traffic, and conversion outcomes to identify response quality gaps and iterate scripts and messaging. Converted real user questions and creator feedback into reusable guidance for better future responses and business performance.• Judged response fit for natural buying intent and persuasive needs.• Performed post-session analysis of how messaging affected engagement and conversion.• Iterated language and scripts in real time based on audience reactions.• Produced evaluation notes from practical livestream interactions and outcomes.

2023 - 2024

Cross-Border E-commerce Operations — Multilingual localized response refinement and evaluation (Medole Co., Ltd.)

Don't discloseText

Localized product information and messaging for Japanese consumers and structured content to fit market-specific search intent and expectations. Reviewed multilingual product data and promotional copy to improve onsite/offsite promotion effectiveness and user satisfaction. Turned localized responses into repeatable patterns for bilingual content handling and cross-cultural communication.• Localized product selling points and customer-facing content for Japan.• Applied search intent analysis to choose keywords and promotional messaging.• Reviewed and refined bilingual product information and translations.• Supported store optimization via data-driven content and localization decisions.

2021 - 2023

Education

Work History