Son N.

Key Skills

Software

Remotasks

Other

Scale AI

Appen

Data Annotation Tech

Top Subject Matter

Cross-lingual AI localization (Kiswahili to English) for NLP datasets

Multimodal audio/video annotation with chronological alignment

Computer vision labeling for object detection and segmentation

Top Data Types

Text

Audio

Image

Document

Video

Top Task Types

Transcription

Segmentation

Data Collection

Freelancer Overview

Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Natural Language Processing (NLP) Alignment, Advanced Multimodal Auditing, and High-Precision Image & Video Frame Labeling powered by extensive platform execution across Remotasks (Scale AI), CrowdGen, and Atlas. Education includes Bachelor of Science, Kirinyaga University School of Pure and Applied Sciences (2023). AI-training focus includes data types such as Text, Audio, and Video and labeling workflows including Translation, Localization, and Evaluation.

Labeling Experience

Language Translation Specialist & AI Localization Consultant (RWS Group)

Text

Provided bidirectional Kiswahili↔English translations and localization workflows for NLP training datasets with high-fidelity accuracy checks. Evaluated semantic structures, idioms, and contextual syntactic patterns to improve how models learn localized cultural nuances. Enforced strict verification and quality assurance metrics to prevent syntax and structural discrepancies across thousands of localized entries. • Localized multilingual training dataset content (Kiswahili to English, and back). • Performed nuanced semantic and contextual evaluation to support model training. • Conducted QA to eliminate contradictions and structural discrepancies in dataset records. • Maintained consistency against localization guidelines for large-scale dataset ingestion.

2025 - Present

Language Translation & AI Evaluation Contractor (CrowdGen)

RemotasksText

Performed targeted cross-lingual translation, localized text generation, and comparative evaluation tasks to supply training pipelines. Reviewed, graded, and refined multilingual model responses for syntactic naturalness, grammatical alignment, and informational accuracy. Used linguistic guidelines to support optimization of tokenizer parsing for East African dialects. • Cross-lingual translation and localized generation for Kiswahili↔English training needs. • Comparative evaluation and grading of multilingual responses for quality dimensions. • Refined outputs to improve alignment with linguistic guidelines. • Supported tokenizer parsing optimization for East African dialects.

2024 - 2025

Video & Audio Transcription Specialist (Atlas AI Data Operations)

AudioTranscription

Converted audio and video media into clean, time-series transcripts for deep learning speech-recognition training. Audited chronological alignment and engineered speaker diarization tags while logging dense non-verbal cues with millisecond accuracy. Cross-referenced transcripts against linguistic quality metrics to remove parsing bugs and optimize acoustic training blocks. • Produced time-series transcripts from multimedia sources. • Added speaker diarization tags and captured non-verbal contextual cues. • Verified chronological alignment with strict timing accuracy. • Audited transcript quality to improve training readiness and reduce parsing errors.

2024 - 2024

Image & Video Labeling Specialist (Computer Vision) (Scale AI / Remotasks Platform)

Scale AIVideoSegmentation

Performed high-precision image and video frame annotation for computer vision training. Implemented 2D/3D bounding boxes, semantic segmentation, and chronological action labeling across video tracks. Audited temporal transitions and spatial occlusions to maintain ground-truth validation above 98% and reported systemic rendering anomalies that impacted label ingestion. • Annotated frames with bounding boxes and semantic segmentation. • Labeled chronological actions across video sequences. • Audited occlusions and temporal transitions for high validation score. • Identified and reported label ingestion/rendering pipeline anomalies to streamline processing.

2023 - 2024

Education

Work History