Ebube Treasure J.

Key Skills

Software

Scale AI

OneForma

Clickworker

Appen

CrowdSource

Top Subject Matter

Audio transcription and training dataset quality control

Speaker diarization and multilingual voice recognition labeling

Content safety evaluation and LLM output grounding assessment

Top Data Types

Audio

Text

Video

Top Task Types

Transcription

Segmentation

Classification

Data Collection

Text Summarization

Freelancer Overview

Data Transcription Specialist (Remote) at ScaleAI. Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Scale AI and OneForma. Education includes Bachelor of Engineering, Nnamdi Azikiwe University (2025) and Level Certificate, Federal Science and Technical College (2019). AI-training focus includes data types such as Audio and Text and labeling workflows including Transcription, Segmentation, and Evaluation.

Labeling Experience

Data Transcription Specialist (Remote) at ScaleAI

Scale AIAudioTranscription

Transcribed 100+ audio files remotely with high accuracy while checking for semantic errors in each dataset. Preserved original context and source integrity to support reliable training data. Applied consistent quality standards and flagged edge cases for reviewer attention to improve dataset structure and acceptance rates. • Transcription of 100+ audio files for ML dataset creation • Semantic error identification while maintaining context • Quality control through consistent guidelines and edge-case flagging • Support of reliable training datasets for downstream clients

2025 - 2026

Content Rater (Remote) at LightHouse Project – OneForma

OneFormaText

Reviewed and flagged AI-generated content across multiple hate speech categories to support safer, more policy-compliant model outputs. Evaluated hundreds of model responses against source texts to identify grounding failures. These checks improved factual accuracy and reduced hallucinations in LLM outputs. • Hate speech category review and flagging • Grounding failure detection against source texts • Evaluation of hundreds of model responses • Improvement of factuality and hallucination reduction via feedback

2025 - 2025

Audio Annotator (Remote) at Polly Project – OneForma

OneFormaAudioSegmentation

Annotated speaker identities across short, predominantly non-English audio files to generate labeled data for voice recognition. Segmented long-form recordings by speaker and annotated sentence-level start and end points with precision. Produced cleaner training data for speaker diarization models to improve multilingual recognition performance. • Speaker identity annotation for non-English audio • Sentence-level start/end timestamp labeling • Speaker-based segmentation for long recordings • Labeled data creation to improve voice recognition accuracy

2025 - 2025

Education

Work History