Connor Smith

Key Skills

Software

Anno-Mage

Appen

CrowdSource

Data Annotation Tech

LabelImg

Labelbox

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Image

Text

Video

Top Label Types

Action Recognition

Audio Recording

Bounding Box

Classification

Computer Programming Coding

Cuboid

Data Collection

Diagnosis

Emotion Recognition

Entity Ner Classification

Evaluation Rating

Fine Tuning

Function Calling

Geocoding

Land Cover Classification

Mapping

Text Summarization

Tracking

Transcription

Freelancer Overview

I am an AI training and data annotation specialist with over 9 years of experience supporting large-scale NLP and LLM development initiatives. My background combines a strong academic foundation in computer science and data science with hands-on expertise in supervised labeling, RLHF, dataset curation, and model evaluation. I have led distributed teams in designing annotation guidelines, managing quality assurance, and implementing bias mitigation strategies, resulting in measurable improvements such as reducing model hallucination rates and increasing annotation reliability. My technical skills include Python, TensorFlow, PyTorch, SQL, and cloud-based annotation platforms, and I am adept at building scalable workflows for prompt engineering, semantic labeling, and inter-annotator agreement analysis. I thrive in remote, cross-functional environments and am passionate about improving model alignment, reasoning quality, and ethical AI through structured human feedback and rigorous data governance.

Labeling Experience

Multimodal AI Training & High-Precision Data Annotation for LLM and Computer Vision

AppenImageBounding BoxPoint Key Point

Led and executed large-scale multimodal data labeling initiatives supporting the training and fine-tuning of advanced AI and large language models. The project involved end-to-end annotation across text, image, and audio datasets to improve model accuracy, contextual understanding, and safety alignment. Scope of Work: Annotated and validated 500,000+ data points across multimodal datasets. Performed high-precision image labeling including bounding boxes, polygon segmentation, object detection, and visual classification. Executed audio annotation workflows such as speech-to-text transcription, speaker diarization, emotion tagging, and acoustic event detection. Conducted NER tagging and text classification for NLP pipelines supporting conversational AI systems. Delivered RLHF and evaluation/rating tasks to improve LLM response quality, factuality, and safety compliance. Produced prompt-response pairs (SFT) to strengthen model instruction-following behavior.

Led and executed large-scale multimodal data labeling initiatives supporting the training and fine-tuning of advanced AI and large language models. The project involved end-to-end annotation across text, image, and audio datasets to improve model accuracy, contextual understanding, and safety alignment. Scope of Work: Annotated and validated 500,000+ data points across multimodal datasets. Performed high-precision image labeling including bounding boxes, polygon segmentation, object detection, and visual classification. Executed audio annotation workflows such as speech-to-text transcription, speaker diarization, emotion tagging, and acoustic event detection. Conducted NER tagging and text classification for NLP pipelines supporting conversational AI systems. Delivered RLHF and evaluation/rating tasks to improve LLM response quality, factuality, and safety compliance. Produced prompt-response pairs (SFT) to strengthen model instruction-following behavior.

2021

Audio Annotation for Speech Recognition & Emotion Detection

Scale AIAudioBounding BoxClassification

Worked on annotating over 5,000 hours of audio recordings for speech recognition and emotion detection projects. Tasks included transcribing spoken content, labeling speaker emotions (e.g., happy, frustrated, neutral), identifying speaker turns, and classifying audio segments by context (customer support, general conversation, commands). Ensured high-quality annotations by following strict labeling guidelines, reviewing annotations for accuracy, and maintaining consistency across datasets to support AI model training.

Worked on annotating over 5,000 hours of audio recordings for speech recognition and emotion detection projects. Tasks included transcribing spoken content, labeling speaker emotions (e.g., happy, frustrated, neutral), identifying speaker turns, and classifying audio segments by context (customer support, general conversation, commands). Ensured high-quality annotations by following strict labeling guidelines, reviewing annotations for accuracy, and maintaining consistency across datasets to support AI model training.

2023 - 2024

Senior Video Annotation Specialist – Multi-Object Tracking & Action Recognition

LabelboxVideoTracking

Executed high-precision video annotation workflows for computer vision model training across multi-domain datasets. Responsible for frame-by-frame object detection, multi-object tracking (MOT), and action recognition in complex real-world video environments. Key responsibilities included: Annotated thousands of video frames using bounding boxes, polygons, and key points Performed temporal tracking of moving objects across long video sequences Labeled human actions and behavioral events for activity recognition models Maintained strict annotation consistency across frames and edge cases Applied occlusion handling, motion blur correction, and ID continuity best practices Conducted quality assurance reviews to maintain 95%+ annotation accuracy Optimized labeling speed while preserving pixel-level precision.

Executed high-precision video annotation workflows for computer vision model training across multi-domain datasets. Responsible for frame-by-frame object detection, multi-object tracking (MOT), and action recognition in complex real-world video environments. Key responsibilities included: Annotated thousands of video frames using bounding boxes, polygons, and key points Performed temporal tracking of moving objects across long video sequences Labeled human actions and behavioral events for activity recognition models Maintained strict annotation consistency across frames and edge cases Applied occlusion handling, motion blur correction, and ID continuity best practices Conducted quality assurance reviews to maintain 95%+ annotation accuracy Optimized labeling speed while preserving pixel-level precision.

2023 - 2024

Education

Work History