Tan Dat Chau

Key Skills

Software

Other

Labelbox

Top Subject Matter

No subject matter listed

Top Data Types

Text

Audio

Document

Video

Top Label Types

Classification

Text Generation

Emotion Recognition

RLHF

Evaluation Rating

Transcription

Data Collection

Prompt Response Writing SFT

Question Answering

Audio Recording

Freelancer Overview

I have hands-on experience contributing to 11+ AI training and data labelling projects across multimodal domains, including text, audio, video, and finance. My work involves evaluating AI-generated outputs for grammar accuracy, conversational smoothness, instruction adherence, and cultural alignment in both English and Vietnamese. I have also designed complex prompts, generated test cases, and developed detailed evaluation rubrics to assess model reasoning capability in the Asset Management context. Beyond traditional annotation, I have also transcribed multi-speaker and aviation (ATC) audio under strict standards, designed visually appealing finance slides, and generated high-quality audio/video datasets to enable model training. With a strong academic background in Applied Finance (Dean’s Medal recipient) and current Master of Finance studies, I am able to onboard seamlessly and contribute effectively to AI projects involving finance. In my past projects, I have performed expert-level financial reasoning tasks that replicate end-to-end asset management analyst workflows. This includes designing complex prompts, providing financial documents, constructing multi-dimensional evaluation rubrics, and evaluating model reasoning power across multi-step analytical tasks. Additionally, being trilingual enables me to work on a wide range of projects across different languages. My combined expertise in finance, multilingual communication, and structured evaluation enables me to create challenging, high-quality training data that enhances model alignment, reasoning accuracy, and real-world applicability.

Labeling Experience

Terminator

LabelboxVideoClassificationQuestion Answering

Captured head-mounted, first-person (POV) video data of human activities in realistic residential environments or non-residential environments to support multimodal model training. Ensured natural task execution and adherence to written collection guidelines to generate high-quality egocentric datasets for activity recognition and behavioral modeling systems.

Captured head-mounted, first-person (POV) video data of human activities in realistic residential environments or non-residential environments to support multimodal model training. Ensured natural task execution and adherence to written collection guidelines to generate high-quality egocentric datasets for activity recognition and behavioral modeling systems.

2025 - 2025

ATC Transcription

LabelboxAudioTranscription

Transcribed ATC audio recordings into precise, well-formatted text in accordance with FAA communication standards. Ensured high accuracy in aviation terminology, call signs, and procedural phraseology to support evaluation and training of speech-recognition models in safety-critical communication environments.

Transcribed ATC audio recordings into precise, well-formatted text in accordance with FAA communication standards. Ensured high accuracy in aviation terminology, call signs, and procedural phraseology to support evaluation and training of speech-recognition models in safety-critical communication environments.

2025

AIDO Finance

LabelboxTextText GenerationRLHF

Generated expert-level financial training data replicating end-to-end analyst workflows within Asset Management companies such as BlackRock, Vanguard, and Fidelity Investments. Designed complex, real-world prompts with supporting input files (e.g., financial statements, market data, portfolio reports) to challenge model reasoning across multi-step analytical tasks. Developed structured evaluation rubrics (critical, major, minor criteria) with 8+ detailed assessment dimensions per task, and systematically evaluated model outputs to identify reasoning gaps, workflow failures, and domain inaccuracies.

Generated expert-level financial training data replicating end-to-end analyst workflows within Asset Management companies such as BlackRock, Vanguard, and Fidelity Investments. Designed complex, real-world prompts with supporting input files (e.g., financial statements, market data, portfolio reports) to challenge model reasoning across multi-step analytical tasks. Developed structured evaluation rubrics (critical, major, minor criteria) with 8+ detailed assessment dimensions per task, and systematically evaluated model outputs to identify reasoning gaps, workflow failures, and domain inaccuracies.

2025

Perfect Slides English Finance

LabelboxDocumentData CollectionPrompt Response Writing SFT

Created realistic user prompts and designed professional, high-quality, and aesthetically pleasing business/educational presentation to construct a dataset which serves as "golden examples" for AI models. Ensured factual accuracy, visual clarity, structured storytelling, and executive-level polish.

Created realistic user prompts and designed professional, high-quality, and aesthetically pleasing business/educational presentation to construct a dataset which serves as "golden examples" for AI models. Ensured factual accuracy, visual clarity, structured storytelling, and executive-level polish.

2025 - 2025

Automotive AI

LabelboxAudioAudio Recording

Recorded provided text prompts in controlled in-car environments (parked and moving conditions) using smartphone microphones to capture natural background noise dynamics. Followed recording guidelines to generate high-quality audio datasets for evaluating speech-recognition model performance under real-world conditions.

Recorded provided text prompts in controlled in-car environments (parked and moving conditions) using smartphone microphones to capture natural background noise dynamics. Followed recording guidelines to generate high-quality audio datasets for evaluating speech-recognition model performance under real-world conditions.

2025 - 2025

Show more

Education

Work History