Yinan Liu

Key Skills

Software

Labelbox

Scale AI

Other

Top Subject Matter

Mandarin Speech Recognition and Conversational AI

Language Model Evaluation and Prompt Engineering

Data Annotation Quality Assurance

Top Data Types

Audio

Document

Image

Text

Top Task Types

Audio Recording

Evaluation Rating

Text Generation

Transcription

Freelancer Overview

Mandarin Advanced AI Data Trainer. Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal, Proprietary Tooling, and Other. AI-training focus includes data types such as Audio and Text and labeling workflows including Transcription, Evaluation, and Rating.

Labeling Experience

xylophone glass

Scale AIAudioAudio RecordingTranscription

This project focused on collecting high-quality Chinese speech recordings for AI training. My main responsibilities included writing and localizing prompts to ensure natural, diverse, and contextually accurate speech samples. The tasks required strong language intuition and creativity to produce prompts that captured different tones, scenarios, and speaking styles. The collected recordings were used to enhance speech recognition and voice synthesis models for multilingual applications.

This project focused on collecting high-quality Chinese speech recordings for AI training. My main responsibilities included writing and localizing prompts to ensure natural, diverse, and contextually accurate speech samples. The tasks required strong language intuition and creativity to produce prompts that captured different tones, scenarios, and speaking styles. The collected recordings were used to enhance speech recognition and voice synthesis models for multilingual applications.

2025 - Present

Cyber Evals

Scale AITextEvaluation Rating

This project involved comparing and rating pairs of Chinese model responses to evaluate the quality, accuracy, and naturalness of each output. The goal was to identify which model produced the more contextually appropriate and human-like answer. My responsibilities included analyzing semantic clarity, fluency, and alignment with user intent, and providing consistent scoring based on detailed rubrics.

This project involved comparing and rating pairs of Chinese model responses to evaluate the quality, accuracy, and naturalness of each output. The goal was to identify which model produced the more contextually appropriate and human-like answer. My responsibilities included analyzing semantic clarity, fluency, and alignment with user intent, and providing consistent scoring based on detailed rubrics.

2024 - Present

Hippocratic

OtherAudioAudio Recording

This project focused on improving the AI model’s ability to recognize and interpret human speech responses in healthcare-related contexts. My initial responsibilities included transcribing and evaluating speech recordings to ensure accuracy and natural expression. As the project progressed, I was promoted to QA (Quality Assurance), where I reviewed other labelers’ work, provided feedback, and ensured strict compliance with project guidelines. My efforts contributed to enhancing the model’s speech recognition accuracy and linguistic understanding, particularly in medical and conversational scenarios requiring high precision.

This project focused on improving the AI model’s ability to recognize and interpret human speech responses in healthcare-related contexts. My initial responsibilities included transcribing and evaluating speech recordings to ensure accuracy and natural expression. As the project progressed, I was promoted to QA (Quality Assurance), where I reviewed other labelers’ work, provided feedback, and ensured strict compliance with project guidelines. My efforts contributed to enhancing the model’s speech recognition accuracy and linguistic understanding, particularly in medical and conversational scenarios requiring high precision.

2025 - 2025

Fishbowl

Scale AITextRLHFEvaluation Rating

This project focused on improving the response quality of the Grok conversational model. My primary responsibilities included writing prompts, evaluating model outputs, and comparing Grok’s responses against competitor models to assess performance differences. I also contributed to developing and refining rubrics to guide consistent and high-quality model evaluations. The project ran for several months and required deep understanding of instruction tuning, prompt design, and linguistic precision to enhance overall model alignment and response accuracy.

This project focused on improving the response quality of the Grok conversational model. My primary responsibilities included writing prompts, evaluating model outputs, and comparing Grok’s responses against competitor models to assess performance differences. I also contributed to developing and refining rubrics to guide consistent and high-quality model evaluations. The project ran for several months and required deep understanding of instruction tuning, prompt design, and linguistic precision to enhance overall model alignment and response accuracy.

2025 - 2025

Education

Work History