Troy Bowen

Key Skills

Software

Appen

Telus

Don't disclose

Top Subject Matter

Large Language Models

Safety Domain Expertise

Trust & Safety

Top Data Types

Text

Image

Video

Top Task Types

RLHF

Segmentation

Entity Ner Classification

Red Teaming

Classification

Freelancer Overview

Lead AI Content Evaluator. Core strengths include Appen, Telus, and Invisible Technologies. Education includes Bachelor of Science, University of Illinois Chicago (2018). AI-training focus includes data types such as Text, Image, and Video and labeling workflows including RLHF, Segmentation, and Entity (NER) Classification.

Labeling Experience

Lead Adversarial Tester

Don T DiscloseTextRed Teaming

I spent this period designing sophisticated ‘jailbreak’ scenarios for LLM adversarial safety tests. I specialized in multi-turn dialogue testing sessions to probe for safety policy violations, notably around PII and dangerous instructions. These sessions often lasted up to 90 minutes per scenario. • Identified and documented model vulnerabilities • Explored multi-dimensional prompt techniques • Pioneered adversarial red teaming best practices • Informed design of model safety guardrails.

I spent this period designing sophisticated ‘jailbreak’ scenarios for LLM adversarial safety tests. I specialized in multi-turn dialogue testing sessions to probe for safety policy violations, notably around PII and dangerous instructions. These sessions often lasted up to 90 minutes per scenario. • Identified and documented model vulnerabilities • Explored multi-dimensional prompt techniques • Pioneered adversarial red teaming best practices • Informed design of model safety guardrails.

2023 - 2025

Lead AI Content Evaluator

AppenTextRLHF

I specialized in Red Teaming and adversarial prompting to uncover model safety risks and vulnerabilities. I performed detailed Evals, ranking model responses for truthfulness and adherence to brand guidelines. I wrote high-quality, long-form reference texts for model fine-tuning. • Conducted dataset audits and provided corrective annotations • Served as subject matter expert for Trust & Safety tasks • Developed ‘ground truth' data for model tuning • Applied multi-dimensional quality criteria for ranking responses.

I specialized in Red Teaming and adversarial prompting to uncover model safety risks and vulnerabilities. I performed detailed Evals, ranking model responses for truthfulness and adherence to brand guidelines. I wrote high-quality, long-form reference texts for model fine-tuning. • Conducted dataset audits and provided corrective annotations • Served as subject matter expert for Trust & Safety tasks • Developed ‘ground truth' data for model tuning • Applied multi-dimensional quality criteria for ranking responses.

2023 - 2025

Senior Fact-Checker & Evaluator

Don T DiscloseText

I audited AI-generated technical documentation for factuality and provided gold standard corrections. Each claim was verified against academic journals and primary sources to retrain the model’s citation engine. My meticulous fact-checking was vital for improving LLM accuracy and reliability. • Conducted technical and academic claim verification • Supplied correction data for retraining purposes • Enhanced model citation accuracy through curated feedback • Provided high-quality gold standard dataset annotations.

I audited AI-generated technical documentation for factuality and provided gold standard corrections. Each claim was verified against academic journals and primary sources to retrain the model’s citation engine. My meticulous fact-checking was vital for improving LLM accuracy and reliability. • Conducted technical and academic claim verification • Supplied correction data for retraining purposes • Enhanced model citation accuracy through curated feedback • Provided high-quality gold standard dataset annotations.

2022 - 2023

AI Training Data Specialist

TelusImageSegmentation

I focused on image and video annotation, labeling frame-by-frame data for computer vision models and ensuring accuracy. I participated in RLHF projects, evaluating helpfulness and harmlessness tradeoffs in AI responses. I refined annotation taxonomies for complex concepts. • Managed high-volume workloads with strict quality control • Labeled multimodal datasets to enhance model training • Utilized specialized annotation tools for computer vision • Provided iterative feedback to technical teams.

I focused on image and video annotation, labeling frame-by-frame data for computer vision models and ensuring accuracy. I participated in RLHF projects, evaluating helpfulness and harmlessness tradeoffs in AI responses. I refined annotation taxonomies for complex concepts. • Managed high-volume workloads with strict quality control • Labeled multimodal datasets to enhance model training • Utilized specialized annotation tools for computer vision • Provided iterative feedback to technical teams.

2021 - 2023

Multimodal Data Specialist

Don T DiscloseVideoSegmentation

I performed frame-by-frame temporal annotation on over 500 hours of video content as part of a multimodal safety audit. I was responsible for identifying borderline content that required nuanced human judgment, such as distinguishing medical procedures from prohibited graphic material. My contributions ensured data integrity and compliance with safety standards. • Executed large-scale temporal event tagging • Maintained strict consistency across safety annotation tasks • Improved visual dataset integrity for model training • Provided human judgment for challenging content edge cases.

I performed frame-by-frame temporal annotation on over 500 hours of video content as part of a multimodal safety audit. I was responsible for identifying borderline content that required nuanced human judgment, such as distinguishing medical procedures from prohibited graphic material. My contributions ensured data integrity and compliance with safety standards. • Executed large-scale temporal event tagging • Maintained strict consistency across safety annotation tasks • Improved visual dataset integrity for model training • Provided human judgment for challenging content edge cases.

2020 - 2022

Show more

Education

Work History