Po Wing Tsang

Key Skills

Software

Appen

Clickworker

Data Annotation Tech

Label Studio

OneForma

Remotasks

Scale AI

Toloka

Telus

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Geospatial Tiled Imagery

Image

Text

Top Task Types

Classification

Evaluation Rating

Mapping

Relationship

RLHF

Freelancer Overview

I have over five years of specialized experience in data labeling and AI training, with hands-on involvement in high-impact projects spanning map data annotation, search evaluation, and large language model (LLM) ranking. My work has included verifying and annotating map features such as points of interest, street addresses, and routes, consistently ensuring data quality by referencing authoritative sources and adhering to rigorous validation protocols. I have also played a key role in evaluating AI-generated text, assessing linguistic naturalness, identifying error categories, ranking model responses, and providing feedback for model improvement. My qualifications include a strong attention to detail, excellent research skills, and proven reliability in delivering accurate, nuanced annotations on large datasets. I consistently maintain high standards through regular peer reviews, golden task assessments, and strict compliance with project guidelines. My ability to handle diverse data types and contribute to both mapping and AI language projects demonstrates my versatility and commitment to data excellence.

Labeling Experience

Advance AI Trainer

Scale AITextRLHFEvaluation Rating

Project Scope: Participated in a LLM evaluation project focused on assessing and comparing the performance of different language models in generating human-like text responses. The primary goal was to evaluate model-generated responses to a diverse set of prompts, focusing on linguistic naturalness and error categorization. Data Labeling: -Assessed the naturalness and conversational quality of AI-generated responses for each prompt. -Ranked the responses from different models for each prompt, providing detailed justifications and qualitative feedback for the assigned ratings. Project Size: -Evaluated responses to 20,000 prompts across different LLM. Quality Measures: -Utilized human review to maintain high annotation accuracy and ensure nuanced evaluation beyond automated metrics.

Project Scope: Participated in a LLM evaluation project focused on assessing and comparing the performance of different language models in generating human-like text responses. The primary goal was to evaluate model-generated responses to a diverse set of prompts, focusing on linguistic naturalness and error categorization. Data Labeling: -Assessed the naturalness and conversational quality of AI-generated responses for each prompt. -Ranked the responses from different models for each prompt, providing detailed justifications and qualitative feedback for the assigned ratings. Project Size: -Evaluated responses to 20,000 prompts across different LLM. Quality Measures: -Utilized human review to maintain high annotation accuracy and ensure nuanced evaluation beyond automated metrics.

2021 - 2024

Advanced AI Data Trainer

Internal Proprietary ToolingTextRLHF

Performed complex Japanese transcription and annotation tasks with over 97% accuracy, ensuring data quality and consistency.

Performed complex Japanese transcription and annotation tasks with over 97% accuracy, ensuring data quality and consistency.

2024 - 2025

Data Analyst

TelusGeospatial Tiled ImagerySegmentationClassification

Project Scope: Worked on a large-scale Maps Search Evaluation project with a focus on data quality for points of interest (POIs). The primary objective was to ensure that map search results and route recommendations provided users with the most accurate, up-to-date, and reliable information. Data Labeling: -Verified the accuracy of POI details such as names, addresses, and pin locations against official sources Checked for outdated or incorrect map entries and flagged them for correction. Project Size: -Evaluated and annotated thousands of POIs and routes across multiple regions and city zones. -Collaborated with a team of raters to systematically review a large volume of map listings and route suggestions. Quality Measures: -Relied on official and authoritative resources for POI validation to ensure data reliability. -“Golden task” quality monitoring, wherein select benchmark tasks were used to continually check and calibrate annotator accuracy.

Project Scope: Worked on a large-scale Maps Search Evaluation project with a focus on data quality for points of interest (POIs). The primary objective was to ensure that map search results and route recommendations provided users with the most accurate, up-to-date, and reliable information. Data Labeling: -Verified the accuracy of POI details such as names, addresses, and pin locations against official sources Checked for outdated or incorrect map entries and flagged them for correction. Project Size: -Evaluated and annotated thousands of POIs and routes across multiple regions and city zones. -Collaborated with a team of raters to systematically review a large volume of map listings and route suggestions. Quality Measures: -Relied on official and authoritative resources for POI validation to ensure data reliability. -“Golden task” quality monitoring, wherein select benchmark tasks were used to continually check and calibrate annotator accuracy.

2019 - 2024

Education

Work History