Data Annotator & Labeling Specialist
Image
Object DetectionData Annotation & Quality Assurance: Annotated large-scale datasets (text, images, and audio) to train machine learning and generative AI models, consistently maintaining an accuracy rating above 98%. RLHF & Prompt Evaluation: Evaluated AI model responses using Reinforcement Learning from Human Feedback (RLHF) frameworks, scoring outputs for factual accuracy, relevance, safety, and grammar. Guideline Adherence: Strictly followed complex, evolving project guidelines to ensure high-quality data ingestion, reducing dataset errors and re-work cycles. Domain Expertise: Categorized text, performed named entity recognition (NER), and wrote high-quality prompts/responses to fine-tune Large Language Models (LLMs). Content Evaluation: Analyzed user-generated content and textual information against strict compliance and quality guidelines, identifying anomalies and correcting errors swiftly. Analytical Thinking: Leveraged strong linguistic and logical reasoning skills to evaluate complex information, translating project requirements into highly accurate outputs. Efficiency & Self-Management: Successfully managed independent workflows in a remote environment, consistently meeting tight daily production targets and quality benchmarks. Edge-Case Resolution & Ambiguity Handling: Identified, documented, and resolved ambiguous data points and complex edge cases not explicitly covered in initial rubrics, establishing precedents that improved overall team labeling consistency. Cross-Functional Feedback Loops: Partnered closely with machine learning engineers and project managers to provide qualitative feedback on model blind spots, directly contributing to the iterative refinement of prompt guidelines. Multi-Turn Conversation Modeling: Authored and evaluated multiturn conversational datasets to train AI chatbots, ensuring the models maintained context, proper tone, and persona consistency over extended interactions. Red Teaming & Safety Testing: Executed adversarial testing (red teaming) by crafting nuanced prompts designed to test the model's safety guardrails, helping to identify and mitigate potential biases, hallucinations, and harmful outputs. Annotation Tool Proficiency: Utilized a variety of proprietary and industry-standard data labeling platforms and UI tools, quickly adapting to diverse software interfaces and keyboard-shortcut workflows to maximize hourly throughput.