Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I support AI training pipelines by applying project-specific guidelines across text, audio, and multimodal datasets—helping teams produce accurate, well-structured training data. My work includes careful evaluation and red-teaming, identifying edge cases, and analyzing model failure patterns to improve instruction-following and reduce hallucinations. I also focus on high-quality annotation and QA, including millisecond timestamp alignment for audio/video, speech transcription, and labeling emotional and vocal attributes. I collaborate across enterprise tools and workflows (Jira, Slack, Google Workspace, Excel, and SQL-based interfaces) and provide structured, stakeholder-ready reporting that documents both quality findings and user impact.…I support AI training pipelines by applying project-specific guidelines across text, audio, and multimodal datasets—helping teams produce accurate, well-structured training data. My work includes careful evaluation and red-teaming, identifying edge cases, and analyzing model failure patterns to improve instruction-following and reduce hallucinations. I also focus on high-quality annotation and QA, including millisecond timestamp alignment for audio/video, speech transcription, and labeling emotional and vocal attributes. I collaborate across enterprise tools and workflows (Jira, Slack, Google Workspace, Excel, and SQL-based interfaces) and provide structured, stakeholder-ready reporting that documents both quality findings and user impact.

Nancy Power

Content Creator, Content Producer, Data Annotator, +7





I support AI training pipelines by applying project-specific guidelines across text, audio, and multimodal datasets—helping teams produce accurate, well-structured training data. My work includes careful evaluation and red-teaming, identifying edge cases, and analyzing model failure patterns to improve instruction-following and reduce hallucinations. I also focus on high-quality annotation and QA, including millisecond timestamp alignment for audio/video, speech transcription, and labeling emotional and vocal attributes. I collaborate across enterprise tools and workflows (Jira, Slack, Google Workspace, Excel, and SQL-based interfaces) and provide structured, stakeholder-ready reporting that documents both quality findings and user impact.…I support AI training pipelines by applying project-specific guidelines across text, audio, and multimodal datasets—helping teams produce accurate, well-structured training data. My work includes careful evaluation and red-teaming, identifying edge cases, and analyzing model failure patterns to improve instruction-following and reduce hallucinations. I also focus on high-quality annotation and QA, including millisecond timestamp alignment for audio/video, speech transcription, and labeling emotional and vocal attributes. I collaborate across enterprise tools and workflows (Jira, Slack, Google Workspace, Excel, and SQL-based interfaces) and provide structured, stakeholder-ready reporting that documents both quality findings and user impact.

Available to hire

I support AI training pipelines by applying project-specific guidelines across text, audio, and multimodal datasets—helping teams produce accurate, well-structured training data. My work includes careful evaluation and red-teaming, identifying edge cases, and analyzing model failure patterns to improve instruction-following and reduce hallucinations.

I also focus on high-quality annotation and QA, including millisecond timestamp alignment for audio/video, speech transcription, and labeling emotional and vocal attributes. I collaborate across enterprise tools and workflows (Jira, Slack, Google Workspace, Excel, and SQL-based interfaces) and provide structured, stakeholder-ready reporting that documents both quality findings and user impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Beginner

Work Experience

Core Contributor at Fleet

February 2, 2026 - June 1, 2026

Selected as one of approximately 10 Core Contributors from a workforce of 800+ evaluators, based on output quality, reliability, and depth of analysis • Evaluated AI agents completing multi-step enterprise workflows at 40+ hours/week, working inside simulated environments replicating Salesforce, Workday, QuickBooks, Outlook, BI dashboards, and DBT pipelines • Designed and refined adversarial evaluation prompts across complex multi-step workflows, requiring agents to discover information, follow sequential instructions, make decisions, and correctly apply earlier findings; surfaced failure types spanning hallucination, verifier misalignment, instruction drift, and incomplete action chains • Reviewed model outputs and workflow traces for hallucinations, logical inconsistencies, verifier alignment, and recurring failure patterns • Worked directly with JSON outputs and SQL-based interfaces to assess structured data quality, reasoning accuracy, and instruction adherence • Identified broken navigation, missing actions, weak validation, unrealistic seeded data, and UI gaps that would block successful task completion • Delivered structured QA reports translating technical findings into clear, actionable feedback for model improvement teams • Selected for OpenClaw, Fleet's advanced agentic evaluation project, responsible for designing and building seed worlds: the simulated enterprise environments used to stress-test AI agent behavior across complex, multi-step workflows spanning Salesforce, Workday, QuickBooks, Outlook, BI dashboards, and DBT pipelines • Served on a small pre-launch environment QA team to review simulated enterprise environments before release to the broader annotator workforce, surfacing data integrity issues, broken workflows, unrealistic content, and environment-level bugs that would have compromised evaluation quality downstream

AI Data Evaluator at Telus

January 1, 2026 - Present

Evaluated AI-generated content and model outputs across text and multimodal tasks for quality, accuracy, and safety compliance • Applied rubric-based frameworks to assess instruction-following, coherence, and response usefulness • Flagged edge cases and systemic failure patterns to support ongoing model improvement

Trainor/Quality Auditor at Mercor

August 2, 2025 - Present

Evaluated AI-generated outputs across text, image, audio, and video modalities using comparative ranking and rubric-based auditing • Conducted millisecond-level audio transcription alignment and labelled emotional tone, vocal attributes, and speech quality for AI training datasets • Identified labelling inconsistencies and systemic model errors to improve training data reliability