Jaime Deonarain

Key Skills

Software

Scale AI

Appen

Top Subject Matter

Autonomous Vehicles

Computer Vision

Search Relevance

Top Data Types

Image

Text

3D Sensor

Audio

Document

Top Task Types

Bounding Box

Classification

Segmentation

Transcription

Freelancer Overview

Data Labeling Specialist (Computer Vision Focus). Brings 2+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Scale AI, Appen, and N. Education includes Bachelor of Science, University at Albany, SUNY (2021). AI-training focus includes data types such as Image, Text, and 3D Sensor and labeling workflows including Bounding Box, Classification, and Segmentation.

Labeling Experience

Data Labeling Specialist (Computer Vision Focus)

Scale AIImageBounding Box

As a Data Labeling Specialist at Scale AI, I labeled images and 3D point cloud data for autonomous driving models. I maintained a 98.5% QA pass rate and developed reference guides for boundary consistency. This role required spotting edge cases and upholding high throughput and accuracy standards. • Daily annotation of 200–300 images including 3D sensor data and occluded scenes • Flagged over 400 unusual edge cases for guideline updates • Created visual annotation references for new team members • Ranked in top 10% of labelers over six months for speed and quality

As a Data Labeling Specialist at Scale AI, I labeled images and 3D point cloud data for autonomous driving models. I maintained a 98.5% QA pass rate and developed reference guides for boundary consistency. This role required spotting edge cases and upholding high throughput and accuracy standards. • Daily annotation of 200–300 images including 3D sensor data and occluded scenes • Flagged over 400 unusual edge cases for guideline updates • Created visual annotation references for new team members • Ranked in top 10% of labelers over six months for speed and quality

2022 - Present

Search Relevance Annotator (Text Labeling)

AppenTextClassification

As a Search Relevance Annotator on the Yukon Project with Appen, I labeled and classified search queries for intent, satisfaction, and spam. My agreement rate with gold test sets was consistently high. I contributed to updating annotation guidelines and participated in calibration meetings. • Labeled over 12,000 search queries using a 7-point relevance scale • Maintained a 96% agreement rate with gold standards • Participated in weekly calls to resolve ambiguous cases • Helped refine annotation documentation with practical examples

As a Search Relevance Annotator on the Yukon Project with Appen, I labeled and classified search queries for intent, satisfaction, and spam. My agreement rate with gold test sets was consistently high. I contributed to updating annotation guidelines and participated in calibration meetings. • Labeled over 12,000 search queries using a 7-point relevance scale • Maintained a 96% agreement rate with gold standards • Participated in weekly calls to resolve ambiguous cases • Helped refine annotation documentation with practical examples

2021 - 2022

Junior Data Labeling Assistant (Contract)

TextClassification

As a Junior Data Labeling Assistant at Metro Insights LLC, I labeled chat log data for sentiment and issue category. This involved reviewing and correcting previous annotations for quality. I also developed tracking tools to support team throughput management. • Labeled customer chat logs for sentiment (positive/negative/neutral) • Categorized issues as billing, technical, or general • Reviewed 1,500 records to catch and correct mislabels • Designed an Excel tracker for daily labeling performance

As a Junior Data Labeling Assistant at Metro Insights LLC, I labeled chat log data for sentiment and issue category. This involved reviewing and correcting previous annotations for quality. I also developed tracking tools to support team throughput management. • Labeled customer chat logs for sentiment (positive/negative/neutral) • Categorized issues as billing, technical, or general • Reviewed 1,500 records to catch and correct mislabels • Designed an Excel tracker for daily labeling performance

2020 - 2021

Audio Transcription Cleanup

AudioTranscription

For Audio Transcription Cleanup, I reviewed and corrected 800 short audio clips from call center conversations. The task included removing filler words, marking speakers, and fixing overlapping speech for clarity. Transcript accuracy improved from 89% to 94% after my rework. • Audited and improved transcripts for 800 audio clips • Identified and split overlapping speech into clear speaker turns • Removed verbal fillers to increase accuracy • Applied quality checks to elevate overall transcript standards

For Audio Transcription Cleanup, I reviewed and corrected 800 short audio clips from call center conversations. The task included removing filler words, marking speakers, and fixing overlapping speech for clarity. Transcript accuracy improved from 89% to 94% after my rework. • Audited and improved transcripts for 800 audio clips • Identified and split overlapping speech into clear speaker turns • Removed verbal fillers to increase accuracy • Applied quality checks to elevate overall transcript standards

Not specified

LLM Safety Labeling

Text

In LLM Safety Labeling, I ranked 500 model responses according to helpfulness and harmlessness using a rubric with six edge-case categories. My labeled rankings contributed directly to tuning a chatbot’s reward model. The work focused on nuanced safety, medical, and political question review. • Evaluated model outputs for six distinct safety and edge-case types • Used structured ranking rubrics for fairness and repeatability • Fine-tuned chatbot model behavior based on label outputs • Labeled responses used in reinforcement learning from human feedback

In LLM Safety Labeling, I ranked 500 model responses according to helpfulness and harmlessness using a rubric with six edge-case categories. My labeled rankings contributed directly to tuning a chatbot’s reward model. The work focused on nuanced safety, medical, and political question review. • Evaluated model outputs for six distinct safety and edge-case types • Used structured ranking rubrics for fairness and repeatability • Fine-tuned chatbot model behavior based on label outputs • Labeled responses used in reinforcement learning from human feedback

Not specified

Show more

Education

Work History