Samuel Gachagua

Key Skills

Software

Clickworker

CloudFactory

Data Annotation Tech

Encord

Labelbox

Appen

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Action Recognition

Audio Recording

Bounding Box

Classification

Emotion Recognition

Freelancer Overview

I am a highly accurate and dependable AI data annotation specialist with extensive experience across all major data modalities: image, video, text, audio, and 3D point clouds. Since 2022, I have successfully completed hundreds of projects on leading platforms (OpenTrainAI, Remotasks, Appen, Scale AI, etc.), consistently achieving 98–100% accuracy scores and ranking in the top 5–10% of contributors. My strength lies in quickly mastering new guidelines and tools (CVAT, Labelbox, Scale Nucleus, Encord, SuperAnnotate, Dataloop) while delivering large volumes of annotations under tight deadlines—often 1,000+ objects per day without compromising quality.What sets me apart is my versatility and obsessive attention to edge cases. I am equally comfortable drawing precise bounding boxes and semantic segmentation for autonomous driving datasets, tracking objects across long video sequences, performing detailed NER and relation extraction on complex documents, evaluating LLM responses for safety and instruction-following, or transcribing and labeling audio in English. Clients repeatedly praise my clear communication, proactive clarification questions, and ability to maintain perfect consistency in multi-annotator projects. With a background in Information Technology, stable high-speed internet, and full-time availability (40–50 hours/week), I am ready to contribute immediately to computer vision, e-commerce, generative AI, or any other high-impact AI training initiative.

Labeling Experience

E-Commerce Product Image Annotation & Attribute Tagging

LabelboxImageBounding BoxPoint Key Point

Labeled 1000s of product images (fashion, electronics, home goods) with precise bounding boxes and polygons for object detection and visual search. Added detailed attribute tags (color, material, brand, style, etc.) and keypoint annotation for clothing fit. Maintained 99% accuracy across multiple clients; frequently invited back for new collections and quality audits.

Labeled 1000s of product images (fashion, electronics, home goods) with precise bounding boxes and polygons for object detection and visual search. Added detailed attribute tags (color, material, brand, style, etc.) and keypoint annotation for clothing fit. Maintained 99% accuracy across multiple clients; frequently invited back for new collections and quality audits.

2025

English Audio Transcription & Emotion/Speaker Labeling

AppenAudioEmotion RecognitionAudio Recording

Transcribed and time-stamped 500+ hours of English audio (conversations, podcasts, call center recordings) with verbatim and clean-read versions. Labeled speaker changes, emotion (happy, angry, neutral, etc.), and background noise. Delivered 99%+ transcription accuracy even with heavy accents and overlapping speech.

Transcribed and time-stamped 500+ hours of English audio (conversations, podcasts, call center recordings) with verbatim and clean-read versions. Labeled speaker changes, emotion (happy, angry, neutral, etc.), and background noise. Delivered 99%+ transcription accuracy even with heavy accents and overlapping speech.

2024

Autonomous Driving – Multi-Sensor Annotation Specialist

EncordImageBounding BoxPolygon

Annotated thousands of street scenes from dashcam and LiDAR data for a leading self-driving car project. Tasks included 2D/3D bounding boxes and cuboids for vehicles/pedestrians, semantic segmentation of drivable areas, lane/polyline marking, and frame-by-frame object tracking in long video sequences. Consistently scored 98%+ IoU accuracy and completed up to 800 objects daily

Annotated thousands of street scenes from dashcam and LiDAR data for a leading self-driving car project. Tasks included 2D/3D bounding boxes and cuboids for vehicles/pedestrians, semantic segmentation of drivable areas, lane/polyline marking, and frame-by-frame object tracking in long video sequences. Consistently scored 98%+ IoU accuracy and completed up to 800 objects daily

2025 - 2025

Education

Work History