Lei Z.

Key Skills

Software

Don't disclose

Other

Top Subject Matter

AI 指令/文本标注与对话质检（物流/供应链垂直场景）

中文文本校对、对账与物流/供应链单据领域 Domain Expertise

Top Data Types

Text

Document

Top Task Types

Prompt + Response Writing (SFT)

Freelancer Overview

在 AI 训练数据处理方面，我目前以深度用户身份积累了系统性的实操经验。日常工作中长期并用 Claude（Sonnet 4.6）、DeepSeek V4 Pro 和 Gemini 3.1 Pro 三款主流大模型进行交叉验证，熟悉三者在中文生成质量、逻辑推理一致性、指令遵循准确度上的差异表现，具备从使用者视角识别模型输出中常见问题的能力，包括事实幻觉、逻辑矛盾、格式偏差及指令歧义等。 在数据处理基础能力方面，我有三年结构化数据录入与多表对账的工作经验，长期负责跨供应商的运输台账维护、应收账单核对与月度汇总报表输出，连续三年保持账单零重大差错记录。这一背景使我能够在高重复性、强规范性的标注任务中保持稳定的准确率与交付质量。中文母语，具备基础英文单据阅读能力，可优先承接物流、供应链及通用对话类垂直领域的文本标注与质检任务。

Labeling Experience

AI数据标注员能力实践—中文校对与结构化整理（物流/供应链垂直）

OtherText

Conducts Chinese document review and accuracy verification for professional shipping and customs materials, then structures and整理 outputs for annotation-ready records. Ensures details are captured correctly through meticulous cross-checking of repeated fields and line items across source documents. Supports high-quality labeling and quality assurance by maintaining near-zero tolerance for errors in tabulated and account-related data. • Long-term processing of Chinese logistics documents such as draft slips, confirmation lists, and bills • Accuracy audits and structural整理 for annotation inputs • Strict error-checking habits for repetitive data fields • Produces reliable records to support labeling/QA workflows

2023 - Present

AI数据标注员（兼职）—指令质量评估与对话质检能力实践

Don't discloseTextPrompt + Response Writing (SFT)

Responsible for prompt quality evaluation by identifying ambiguities, hallucinations, logical contradictions, and formatting issues in AI outputs during daily multi-model use. Uses multi-dimensional scoring practices to judge consistency and quality from a user perspective, supporting text and instruction-oriented optimization tasks. Applies iterative cross-checking to ensure high-fidelity labeling and reduced error rates in downstream annotation work. • Multi-model cross validation with Claude (Sonnet 4.6), DeepSeek V4 Pro, and Gemini 3.1 Pro • Prompt quality checks including instruction ambiguity, hallucination detection, and format compliance • Multi-dimensional rating workflow for AI responses • Consistency validation across model outputs for reliable labeling results

2023 - Present

Education

Work History