Crystal Valls

Key Skills

Software

Labelbox

Mercor

Roboflow

Prodigy

Top Subject Matter

No subject matter listed

Top Data Types

Image

Audio

Video

Text

Top Label Types

Bounding Box

Polygon

Segmentation

Classification

Object Detection

Emotion Recognition

Evaluation Rating

Audio Recording

Transcription

Action Recognition

Tracking

Question Answering

Text Generation

RLHF

Prompt Response Writing SFT

Freelancer Overview

I am an AI data specialist with extensive hands-on experience in labeling, annotating, and validating large-scale multimodal datasets across image, audio, text, and video domains. My background in computer science and machine learning allows me to expertly handle complex annotation tasks, including object detection, semantic segmentation, audio event tagging, transcription validation, and NLP data labeling such as sentiment analysis and named entity recognition. I have worked with a wide range of industry-standard tools like Label Studio, CVAT, Supervisely, VGG Image Annotator, and Prodigy, and have annotated over 200,000 data samples to support the development and fine-tuning of advanced AI models. My strengths include building robust data pipelines, developing automated preprocessing scripts in Python, and ensuring dataset quality through rigorous error analysis, bias detection, and quality scoring. I am skilled in prompt engineering, model evaluation, and human-in-the-loop training workflows, contributing to safer and more reliable AI systems. With experience in high-volume environments and a focus on quality assurance, I am passionate about delivering precise, high-quality training data for computer vision, speech recognition, and NLP projects.

Labeling Experience

LLM Evaluation, RLHF and NLP Data annotation project.

ProdigyTextQuestion AnsweringText Generation

Supported training and fine turning of transformer based language model through structured evaluation and reinforcement learning feedback workflows. Annotated and evaluated 200,000+ text samples in,cluding prompts, model outputs and conversational exchanges. Responsibilities included: .Ranking Model responses for coherence and relevance. .Identifying hallucination and factual inaccuracies. .Bias and safety evaluation . .Named entitiy recognition. .Writing high quality supervised fine-turning datay. .Red-teaming model outputs for vulnerability testing.

Supported training and fine turning of transformer based language model through structured evaluation and reinforcement learning feedback workflows. Annotated and evaluated 200,000+ text samples in,cluding prompts, model outputs and conversational exchanges. Responsibilities included: .Ranking Model responses for coherence and relevance. .Identifying hallucination and factual inaccuracies. .Bias and safety evaluation . .Named entitiy recognition. .Writing high quality supervised fine-turning datay. .Red-teaming model outputs for vulnerability testing.

2021

Video Annotation and Frame Tracking for action Recognition Models

RoboflowVideoBounding BoxSegmentation

Annotated large-scale video datataset to support training for action recognition and object tracking models. Key tasks include: .Frame-by-frame Bounding Box Annotation. .Object tracking accross sequences. .Temporaly segmentation of actions. .Scenes classification. .Motion labelling and activity recognition. .Quality validation of annotated frames.

Annotated large-scale video datataset to support training for action recognition and object tracking models. Key tasks include: .Frame-by-frame Bounding Box Annotation. .Object tracking accross sequences. .Temporaly segmentation of actions. .Scenes classification. .Motion labelling and activity recognition. .Quality validation of annotated frames.

2020

Speech and Audio Data Annotation for ASR Model Training

MercorAudioClassificationEmotion Recognition

Annotated and processed large scale speech datasets for automatic speech recognition and conversation AI systems. Responsibilities include: .High-accuracy transcription for diverse speech samples. .Timestamp alignment and segmentation. .Emotional and sentiment labelling. .Audio quality assessment and noise filtering. .Speaker intent classification. .Dataset cleaning and preprosecessing.

Annotated and processed large scale speech datasets for automatic speech recognition and conversation AI systems. Responsibilities include: .High-accuracy transcription for diverse speech samples. .Timestamp alignment and segmentation. .Emotional and sentiment labelling. .Audio quality assessment and noise filtering. .Speaker intent classification. .Dataset cleaning and preprosecessing.

2020

Computer Vision Image Annotation and Object Detection

LabelboxImageBounding BoxPolygon

Performed large scale Image annotation for computer vision model training and validation. Annotated 100,000+ images across diverse datasets including real world scenes,objects and motion based sequences. Tasks included: .Bounding Box and Polygon annotations for object detection. .Semantic and instance segmentation. .Keypoint labelling for pose estimation. .Multi-label classification .Frame by frame tracking for motion based datasets. .Dataset cleaning and annotation validation.

Performed large scale Image annotation for computer vision model training and validation. Annotated 100,000+ images across diverse datasets including real world scenes,objects and motion based sequences. Tasks included: .Bounding Box and Polygon annotations for object detection. .Semantic and instance segmentation. .Keypoint labelling for pose estimation. .Multi-label classification .Frame by frame tracking for motion based datasets. .Dataset cleaning and annotation validation.

2020

Education

Work History