Derek Salinas

Key Skills

Software

Labelbox

CVAT

Appen

Top Subject Matter

No subject matter listed

Top Data Types

Image

Video

Audio

Top Label Types

Polygon

Tracking

Transcription

Bounding Box

Point Key Point

Freelancer Overview

I am an AI training specialist and data annotation expert with over 5 years of hands-on experience providing high-quality labeled data for machine learning models. My expertise spans image, video, and audio annotation, with a strong focus on computer vision and natural language processing applications such as autonomous vehicles, healthcare imaging, retail analytics, and speech recognition systems.

Labeling Experience

Multi-object Video Annotation Specialist

CVATVideoTracking

I led multi-object tracking projects to annotate and analyze customer movement patterns in retail environments. This work involved tracking objects across thousands of hours of in-store video footage, using tools such as CVAT to ensure accuracy and consistency. My contributions improved insights into shopper behaviors and store layout optimization for clients. • Executed tracking on 10,000+ video hours with high precision • Utilized custom protocols for multi-object annotation • Implemented quality assurance to reduce annotation error by 35% • Provided actionable data for retail analytics use cases

I led multi-object tracking projects to annotate and analyze customer movement patterns in retail environments. This work involved tracking objects across thousands of hours of in-store video footage, using tools such as CVAT to ensure accuracy and consistency. My contributions improved insights into shopper behaviors and store layout optimization for clients. • Executed tracking on 10,000+ video hours with high precision • Utilized custom protocols for multi-object annotation • Implemented quality assurance to reduce annotation error by 35% • Provided actionable data for retail analytics use cases

2024 - 2025

Independent AI Training Specialist/Data Annotation Expert

LabelboxImagePolygon

As an independent AI Training Specialist, I managed data annotation pipelines for clients in sectors such as autonomous vehicles, healthcare imaging, and retail analytics. I was responsible for producing annotated image datasets with high precision and inter-annotator agreement. My work included executing bounding box and polygon annotations, leading to improved machine learning model performance. • Managed end-to-end annotation workflows for 15+ clients • Produced 500,000+ labeled images using Labelbox and CVAT • Developed custom YOLO datasets and reduced model false positives by 23% • Collaborated with ML engineers to refine guidelines based on feedback

As an independent AI Training Specialist, I managed data annotation pipelines for clients in sectors such as autonomous vehicles, healthcare imaging, and retail analytics. I was responsible for producing annotated image datasets with high precision and inter-annotator agreement. My work included executing bounding box and polygon annotations, leading to improved machine learning model performance. • Managed end-to-end annotation workflows for 15+ clients • Produced 500,000+ labeled images using Labelbox and CVAT • Developed custom YOLO datasets and reduced model false positives by 23% • Collaborated with ML engineers to refine guidelines based on feedback

2020 - 2022

image

LabelboxImageBounding BoxPolygon

led image annotation operations for autonomous driving perception dataset (350,000 front-camera frames). Managed team of 35 annotators performing 3D cuboid annotation (KITTI format), 2D bounding boxes (12 vehicle/pedestrian classes), and drivable area segmentation (19-class taxonomy). Implemented CVAT-based pipeline with SAM-assisted pre-labeling reducing annotation time by 40%. Enforced strict quality protocol: 0.87 Fleiss' kappa inter-annotator agreement, 0.92 mean IoU against golden set, <5% rejection rate. Delivered COCO-compliant dataset supporting production object detection model achieving 0.78 mAP on COCO evaluation metrics

led image annotation operations for autonomous driving perception dataset (350,000 front-camera frames). Managed team of 35 annotators performing 3D cuboid annotation (KITTI format), 2D bounding boxes (12 vehicle/pedestrian classes), and drivable area segmentation (19-class taxonomy). Implemented CVAT-based pipeline with SAM-assisted pre-labeling reducing annotation time by 40%. Enforced strict quality protocol: 0.87 Fleiss' kappa inter-annotator agreement, 0.92 mean IoU against golden set, <5% rejection rate. Delivered COCO-compliant dataset supporting production object detection model achieving 0.78 mAP on COCO evaluation metrics

2019 - 2020

Audio Transcription & Speaker Diarization Specialist

AppenAudioTranscription

I created and implemented audio transcription protocols for multilingual speech recognition projects. This role required annotating and transcribing over 2,000 hours of audio data, ensuring accurate speaker diarization and segmentation. My work supported the development of advanced voice assistant systems and speech recognition models. • Processed high volumes of multilingual audio using Audacity and Praat • Developed quality standards for consistent transcription • Collaborated with machine learning teams to improve feedback loops • Contributed to improved recognition accuracy across languages

I created and implemented audio transcription protocols for multilingual speech recognition projects. This role required annotating and transcribing over 2,000 hours of audio data, ensuring accurate speaker diarization and segmentation. My work supported the development of advanced voice assistant systems and speech recognition models. • Processed high volumes of multilingual audio using Audacity and Praat • Developed quality standards for consistent transcription • Collaborated with machine learning teams to improve feedback loops • Contributed to improved recognition accuracy across languages

2019 - 2020

Education

Work History