Wasei Ohta

Key Skills

Software

Appen

Clickworker

Data Annotation Tech

Labelbox

Mindrift

OneForma

Remotasks

Scale AI

Toloka

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Document

Text

Top Task Types

Classification

Prompt Response Writing SFT

Red Teaming

RLHF

Freelancer Overview

Bilingual AI Annotator and Quality Team Lead with extensive experience in creating, auditing, and refining datasets for generative AI and RLHF (Reinforcement Learning from Human Feedback) projects. Skilled in red teaming, disclaimer and rejection generation, multimodal and multi-turn dataset creation, and quality audits focused on AI safety, fairness, and controllability. Proven ability to apply detailed evaluation guidelines, detect bias or policy violations, and deliver actionable feedback to improve model reliability and accuracy. In addition to annotation and review expertise, served as Project Coordinator, managing task distribution, onboarding, and communication across a global team of annotators. Specialized in Japanese-English localization and linguistic QA, ensuring natural fluency and cultural consistency in AI-generated content. Experienced in audio transcription labeling, speaker segmentation, and dataset structuring for ASR and conversational AI. Adept at maintaining high-quality standards under tight deadlines and collaborating with international clients to deliver precise, scalable data solutions.

Labeling Experience

Audio Transcription Labeling

LabelboxAudioEvaluation RatingData Collection

Performed precise transcription labeling of Japanese and English audio clips, including speaker identification, timestamping, and text normalization. Ensured phonetic accuracy, consistent labeling conventions, and data alignment for speech-to-text AI model training. Additionally, applied appropriate labeling for non-verbal audio actions—such as laughter, pauses, background noise, and emotional cues—to enhance the model’s understanding of contextual and paralinguistic elements in speech data.

Performed precise transcription labeling of Japanese and English audio clips, including speaker identification, timestamping, and text normalization. Ensured phonetic accuracy, consistent labeling conventions, and data alignment for speech-to-text AI model training. Additionally, applied appropriate labeling for non-verbal audio actions—such as laughter, pauses, background noise, and emotional cues—to enhance the model’s understanding of contextual and paralinguistic elements in speech data.

2025

Audio Recording

Data Annotation TechAudioAudio Recording

Trained and evaluated conversational AI models to ensure appropriate responses were generated from given audio prompts. Reviewed model outputs for accuracy, tone, and contextual relevance; when necessary, revised responses to meet project guidelines. Created high-quality, bilingual training data aligned with defined response standards.

Trained and evaluated conversational AI models to ensure appropriate responses were generated from given audio prompts. Reviewed model outputs for accuracy, tone, and contextual relevance; when necessary, revised responses to meet project guidelines. Created high-quality, bilingual training data aligned with defined response standards.

2025 - 2025

Red Teaming

Data Annotation TechTextEvaluation RatingPrompt Response Writing SFT

Conducted red teaming tasks by generating and reviewing adversarial prompts to identify policy-sensitive or unsafe outputs. Authored natural and compliant disclaimer and rejection responses to ensure model alignment with ethical and safety standards. Managed output control for sensitive or restricted content, ensuring the model responded with appropriate disclaimers or rejections when encountering harmful, biased, or policy-violating inputs. Contributed to improving the model’s ability to handle sensitive topics responsibly and maintain guideline compliance.

Conducted red teaming tasks by generating and reviewing adversarial prompts to identify policy-sensitive or unsafe outputs. Authored natural and compliant disclaimer and rejection responses to ensure model alignment with ethical and safety standards. Managed output control for sensitive or restricted content, ensuring the model responded with appropriate disclaimers or rejections when encountering harmful, biased, or policy-violating inputs. Contributed to improving the model’s ability to handle sensitive topics responsibly and maintain guideline compliance.

2024 - 2025

Multi-modal RLHF

Scale AIImageEvaluation RatingRed Teaming

Recognized objects and visual elements within uploaded images and audited AI-generated responses against corresponding text prompts. Created text prompts based on guideline-compliant images, reviewed the generated responses, and refined them according to rubric standards. Produced high-quality, contextually aligned response data to enhance multimodal AI model accuracy and reasoning performance. Additionally, managed output control for sensitive or policy-violating content by drafting appropriate disclaimers and rejection responses to ensure safety and compliance with project guidelines.

Recognized objects and visual elements within uploaded images and audited AI-generated responses against corresponding text prompts. Created text prompts based on guideline-compliant images, reviewed the generated responses, and refined them according to rubric standards. Produced high-quality, contextually aligned response data to enhance multimodal AI model accuracy and reasoning performance. Additionally, managed output control for sensitive or policy-violating content by drafting appropriate disclaimers and rejection responses to ensure safety and compliance with project guidelines.

2024 - 2025

Prompt/Response Dataset Creation (Single/Multi-turn)

Scale AITextEvaluation RatingPrompt Response Writing SFT

Created and refined single-turn and multi-turn conversational datasets for generative AI models. Focused on ensuring natural flow, factual consistency, and contextual relevance in Japanese. Contributed to improving model instruction-following behavior and reducing ambiguity through iterative prompt optimization.

Created and refined single-turn and multi-turn conversational datasets for generative AI models. Focused on ensuring natural flow, factual consistency, and contextual relevance in Japanese. Contributed to improving model instruction-following behavior and reducing ambiguity through iterative prompt optimization.

2023 - 2025

Education

Work History