Raymond M.

Key Skills

Software

Don't disclose

Other

CVAT

Top Subject Matter

Physical AI model training data (multimodal/video)

LLM preference/rubric-based evaluation for RLHF and safety

Multilingual ASR and RLHF training data

Top Data Types

Video

Text

Audio

Image

Document

Top Task Types

Data Collection

RLHF

Transcription

Tracking

Freelancer Overview

Atlas Capture – AI Data Annotator (Physical AI). Brings 6+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Don't disclose, Other, and CVAT. Education includes Master of Arts, United States International University (2024) and Bachelor of Arts, United States International University (2021). AI-training focus includes data types such as Video, Text, and Audio and labeling workflows including Data Collection, RLHF, and Transcription.

Labeling Experience

Atlas Capture – AI Data Annotator (Physical AI)

Don't discloseVideoData Collection

Provided detailed guideline-driven annotations for video, image, and multimodal datasets supporting Physical AI model training. Applied English annotation guidelines to complex real-world scenarios while maintaining precision, attention to detail, and self-review consistency. Met quality benchmarks for accuracy and completeness while ensuring confidentiality and data security compliance. • Annotated across video, image, and multimodal inputs • Performed self-review for consistency • Escalated ambiguous examples in English • Participated in training and calibration updates

2026 - Present

CVAT.ai – Video Annotation Specialist

CVATVideoTracking

Performed video annotation in CVAT for object detection and tracking tasks including vehicles, pedestrians, and action recognition. Annotated 10,000+ video frames and achieved 99.2% accuracy under high-volume throughput. Developed a QA process using random sampling, error logging, and re-annotation that reduced team error rate by 25% and maintained 99.5% ID consistency while creating training materials for new annotators. • Object detection and tracking on video frames • 10,000+ frames annotated with 99.2% accuracy • QA workflow reduced team error rate by 25% • 5,000 frames labeled with 99.5% ID consistency

2023 - Present

Freelance / Upwork & CrowdGen – Multilingual Data Annotator & Linguist

Don't discloseAudioTranscription

Delivered multilingual AI training data labeling spanning transcription, pairwise evaluation, data collection, and content creation. Transcribed 500+ hours of conversational audio across Swahili, English, Lao, and Amharic with high accuracy to improve ASR models. Conducted RLHF-related data work, including a Swahili-language project that reduced bias toward Western cultural references by 30% using post-deployment metrics. • Multilingual data collection and transcription • Pairwise evaluation and content creation • RLHF bias reduction via targeted labeling • Cross-modal evaluation and tagging across audio/video/images

2021 - Present

Mindrift – AI Trainer & Data Quality Specialist

OtherTextRLHF

Performed LLM evaluation work using pairwise comparisons and gold rubrics across science, finance, marketing, and creative domains. Designed and executed rubrics covering fluency, coherence, factuality, bias, and safety, and flagged hallucinations and subtle misleading statements. Delivered granular written justifications that drove a 25% improvement in model response quality over six months while collaborating with engineers on prompt improvements. • 15,000+ pairwise comparisons of LLM outputs • Rubric evaluation for bias, safety, and factuality • Flagged hallucinations and misleading statements • Provided justification writing supporting quality gains

2024 - 2025

Education

Work History