Oluwayinka Odufote

Key Skills

Software

CVAT

Labelbox

Internal/Proprietary Tooling

Lionbridge

Mindrift

Humanatic

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Document

Video

Audio

Top Label Types

Bounding Box

Segmentation

Text Generation

Object Detection

Classification

RLHF

Data Collection

Fine Tuning

Question Answering

Action Recognition

Prompt Response Writing SFT

Audio Recording

Transcription

Freelancer Overview

I am an experienced Data Annotation Specialist with a strong background in labeling and preparing high-quality datasets for AI model training across image, video, text, and audio domains. My expertise includes creating bounding boxes, caption matching, timestamp labeling, and generating natural language descriptions for computer vision and NLP projects. I have worked extensively with annotation tools like CVAT, Labelbox, and SuperAnnotate, consistently maintaining over 98% accuracy in high-volume, fast-paced remote workflows. My projects have supported LLM training for robotic task recognition, e-commerce product search assistants, and audio transcription for speech recognition and TTS models. I am adept at following detailed guidelines, collaborating with global teams, and ensuring data quality and compliance, and I am eager to contribute my skills to challenging annotation projects that drive AI automation and innovation.

Labeling Experience

Audio Data Quality Assurance & Speech Model Training Support (Inflexion Transcribe)

Internal Proprietary ToolingAudioPrompt Response Writing SFTAudio Recording

Worked as an Audio Data & Annotation Specialist contributing to the training and validation of speech recognition and text-to-speech (TTS) systems. Conducted quality assurance analysis on AI-generated voice reproductions, comparing original source audio against multiple synthesized outputs. Evaluated and selected the closest match based on: Pitch accuracy Vocal tone consistency Timing and speech rhythm Loudness balance Background noise levels Identity similarity to the original speaker Ensured that synthesized outputs met fidelity and naturalness standards required for production-level voice models. Performed high-accuracy transcription of podcast and long-form audio content, converting speech into structured text datasets. Maintained strict standards for: Grammar and punctuation Speaker identification Timestamp accuracy (when required) Verbatim vs. clean transcription formatting

Worked as an Audio Data & Annotation Specialist contributing to the training and validation of speech recognition and text-to-speech (TTS) systems. Conducted quality assurance analysis on AI-generated voice reproductions, comparing original source audio against multiple synthesized outputs. Evaluated and selected the closest match based on: Pitch accuracy Vocal tone consistency Timing and speech rhythm Loudness balance Background noise levels Identity similarity to the original speaker Ensured that synthesized outputs met fidelity and naturalness standards required for production-level voice models. Performed high-accuracy transcription of podcast and long-form audio content, converting speech into structured text datasets. Maintained strict standards for: Grammar and punctuation Speaker identification Timestamp accuracy (when required) Verbatim vs. clean transcription formatting

2025

Video Segmentation & Temporal Action Annotation for AI Model Training

HumanaticVideoSegmentationObject Detection

Worked as a Video Annotation Specialist contributing to computer vision and action-recognition model training. Performed frame-level video segmentation and timestamp-based annotation, breaking down short-form videos into structured time intervals (e.g., 0.0–10.0 seconds). Precisely identified and marked: Start and end of actions Object interactions (e.g., picking up or dropping items) Movement transitions Idle/no-action periods Behavioral changes within the scene Produced concise, objective descriptions of each segment using standardized annotation guidelines to ensure clarity and consistency. Maintained strict temporal accuracy by aligning written descriptions with exact timestamps, enabling AI systems to learn action boundaries, motion patterns, and event sequencing. Contributed to improving model performance in: Action recognition Event detection Human-object interaction modeling Video summarization systems

Worked as a Video Annotation Specialist contributing to computer vision and action-recognition model training. Performed frame-level video segmentation and timestamp-based annotation, breaking down short-form videos into structured time intervals (e.g., 0.0–10.0 seconds). Precisely identified and marked: Start and end of actions Object interactions (e.g., picking up or dropping items) Movement transitions Idle/no-action periods Behavioral changes within the scene Produced concise, objective descriptions of each segment using standardized annotation guidelines to ensure clarity and consistency. Maintained strict temporal accuracy by aligning written descriptions with exact timestamps, enabling AI systems to learn action boundaries, motion patterns, and event sequencing. Contributed to improving model performance in: Action recognition Event detection Human-object interaction modeling Video summarization systems

2025

Human-in-the-Loop AI Model Training & Data Structuring Optimization (Mindrift)

MindriftDocumentSegmentationQuestion Answering

Worked as a Human AI Trainer and Data Quality Specialist supporting machine learning systems through structured evaluation and refinement tasks. Conducted facial image labeling and longitudinal identity verification tasks, uploading historical and recent images to train AI systems in age progression recognition and identity consistency modeling. Helped improve facial recognition algorithms by ensuring accurate person-matching across significant time gaps. Performed AI-assisted research validation and ranking optimization, reviewing machine-generated lists (e.g., SEO influencers, industry leaders) and correcting inaccuracies. Applied human judgment to: Re-rank results based on relevance and authority Verify follower counts and social metrics Cross-check LinkedIn, Instagram, and YouTube profiles Ensure data completeness and credibility Provided structured human oversight to improve AI accuracy in information retrieval, ranking logic, and data compilation integrity.

Worked as a Human AI Trainer and Data Quality Specialist supporting machine learning systems through structured evaluation and refinement tasks. Conducted facial image labeling and longitudinal identity verification tasks, uploading historical and recent images to train AI systems in age progression recognition and identity consistency modeling. Helped improve facial recognition algorithms by ensuring accurate person-matching across significant time gaps. Performed AI-assisted research validation and ranking optimization, reviewing machine-generated lists (e.g., SEO influencers, industry leaders) and correcting inaccuracies. Applied human judgment to: Re-rank results based on relevance and authority Verify follower counts and social metrics Cross-check LinkedIn, Instagram, and YouTube profiles Ensure data completeness and credibility Provided structured human oversight to improve AI accuracy in information retrieval, ranking logic, and data compilation integrity.

2025

Image Evaluator & Search Relevance Optimization (Lionbridge)

LionbridgeImageSegmentationClassification

Served as an AI Image Evaluator and Search Relevance Analyst contributing to the training and optimization of large-scale machine learning systems. As an Image Evaluator, assessed multiple AI-generated images based on pre-written text prompts. Ranked outputs according to realism, natural composition, prompt alignment, and visual accuracy. Selected the most contextually accurate and high-quality images to improve model output performance and reduce synthetic artifacts. As a Search Relevance Evaluator, analyzed user search queries and evaluated product results for functional and contextual relevance. Determined whether returned results were: Exact matches Functional substitutes Complementary products Irrelevant results Applied structured reasoning to ensure that product listings aligned with user intent, mirroring real-world e-commerce evaluation standards. Contributed to improving AI understanding of search intent, product classification accuracy, and result ranking logic.

Served as an AI Image Evaluator and Search Relevance Analyst contributing to the training and optimization of large-scale machine learning systems. As an Image Evaluator, assessed multiple AI-generated images based on pre-written text prompts. Ranked outputs according to realism, natural composition, prompt alignment, and visual accuracy. Selected the most contextually accurate and high-quality images to improve model output performance and reduce synthetic artifacts. As a Search Relevance Evaluator, analyzed user search queries and evaluated product results for functional and contextual relevance. Determined whether returned results were: Exact matches Functional substitutes Complementary products Irrelevant results Applied structured reasoning to ensure that product listings aligned with user intent, mirroring real-world e-commerce evaluation standards. Contributed to improving AI understanding of search intent, product classification accuracy, and result ranking logic.

2025 - 2025

Data Annotation Specialist (AnnotateNow)

CVATImageBounding BoxSegmentation

Labeled and annotated thousands of images and videos for AI model training and computer vision projects focused on object detection and robotic task segmentation. Maintained a 98%+ accuracy rate while segmenting robotic task videos into subtasks using precise timestamp annotation. Supported the annotation of robot-human interaction videos for LLM training on task recognition. • Annotated 500+ videos of robot-human interactions for AI task recognition. • Created and validated bounding boxes for over 1,200 objects weekly. • Segmented videos into subtasks with accurate timestamp labeling. • Evaluated conversational context and speaker emotion during transcription of podcast content.

Labeled and annotated thousands of images and videos for AI model training and computer vision projects focused on object detection and robotic task segmentation. Maintained a 98%+ accuracy rate while segmenting robotic task videos into subtasks using precise timestamp annotation. Supported the annotation of robot-human interaction videos for LLM training on task recognition. • Annotated 500+ videos of robot-human interactions for AI task recognition. • Created and validated bounding boxes for over 1,200 objects weekly. • Segmented videos into subtasks with accurate timestamp labeling. • Evaluated conversational context and speaker emotion during transcription of podcast content.

2024 - 2025

Show more

Education

Work History