Steven F.

Key Skills

Software

Don't disclose

Other

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

Text

Image

Top Task Types

Evaluation Rating

Fine Tuning

Transcription

Entity Ner Classification

Classification

Data Collection

Freelancer Overview

I am a technically driven professional with hands-on experience in data operations, automation testing, and AI-focused projects. My background includes managing large-scale server operations, monitoring data metrics, and ensuring high-quality data reporting in fast-paced environments. I have worked extensively with Python, MATLAB, SQL/NoSQL, AWS, and automation frameworks, applying these skills to improve data accuracy and operational efficiency. Notably, I have researched machine learning algorithms for OCR classification and optimized data training procedures for faster and more precise results. My experience spans QA testing, project management, and cross-functional collaboration, making me adept at supporting high-quality data annotation and AI training workflows across diverse domains. I have hands-on experience training and evaluating LLMs through work with DataAnnotationTech, Mindrift, and Outlier across tasks such as annotation, classification, rewriting, ranking, safety evaluation, and reasoning assessment. My technical background in data operations, QA testing, and ML-related research enables me to analyze complex model outputs, identify inconsistencies, and provide high-precision feedback. I follow guidelines exactly, adapt quickly to new task formats, and consistently deliver accurate, high-quality annotations for AI model improvement.

Labeling Experience

AI Model Response Ranking & Content Quality Evaluation – Outlier

OtherTextFine TuningEvaluation Rating

Evaluated multiple model-generated responses and ranked them based on accuracy, clarity, depth, reasoning, creativity, and adherence to task requirements. Rewrote and improved weak outputs to serve as high-quality reference exemplars. Classified content into categories, flagged errors, and applied detailed rubrics to measure quality. Provided consistent scoring and annotations to help refine model preference alignment and reduce hallucinations.

Evaluated multiple model-generated responses and ranked them based on accuracy, clarity, depth, reasoning, creativity, and adherence to task requirements. Rewrote and improved weak outputs to serve as high-quality reference exemplars. Classified content into categories, flagged errors, and applied detailed rubrics to measure quality. Provided consistent scoring and annotations to help refine model preference alignment and reduce hallucinations.

2025 - 2025

Aether LLM Evaluation & Reasoning Improvement – Outlier

OtherImageFine TuningEvaluation Rating

Participated in the Aether project supporting advanced reasoning evaluation and instruction refinement for LLM improvement. Responsibilities included reviewing multi-step reasoning chains, identifying logical inconsistencies, and ranking alternative model outputs based on accuracy, coherence, and alignment with task constraints. Produced high-quality rewritten responses to serve as gold-standard exemplars, ensuring improved model generalization and chain-of-thought structure. Analyzed complex prompts involving multi-hop logic, mathematics, problem-solving, procedural steps, and long-form explanations. Applied detailed rubrics to validate factual correctness, thought structure, and adherence to safety policies. Delivered consistent, guideline-aligned judgments across thousands of items, contributing to refinement of preference models and next-generation reasoning architectures.

Participated in the Aether project supporting advanced reasoning evaluation and instruction refinement for LLM improvement. Responsibilities included reviewing multi-step reasoning chains, identifying logical inconsistencies, and ranking alternative model outputs based on accuracy, coherence, and alignment with task constraints. Produced high-quality rewritten responses to serve as gold-standard exemplars, ensuring improved model generalization and chain-of-thought structure. Analyzed complex prompts involving multi-hop logic, mathematics, problem-solving, procedural steps, and long-form explanations. Applied detailed rubrics to validate factual correctness, thought structure, and adherence to safety policies. Delivered consistent, guideline-aligned judgments across thousands of items, contributing to refinement of preference models and next-generation reasoning architectures.

2024 - 2025

LLM Output Evaluation & Instruction Refinement – Mindrift

Don T DiscloseTextEvaluation Rating

Reviewed and evaluated LLM-generated responses for correctness, clarity, safety, and alignment with task instructions. Performed multi-step reasoning assessments, comparative ranking, rewriting, and classification of model outputs. Identified inconsistencies, edge cases, and failure patterns, providing structured feedback for improving model performance. Followed detailed guidelines for content safety, logical accuracy, and linguistic quality. Handled large volumes of text-based tasks with high accuracy and consistency.

Reviewed and evaluated LLM-generated responses for correctness, clarity, safety, and alignment with task instructions. Performed multi-step reasoning assessments, comparative ranking, rewriting, and classification of model outputs. Identified inconsistencies, edge cases, and failure patterns, providing structured feedback for improving model performance. Followed detailed guidelines for content safety, logical accuracy, and linguistic quality. Handled large volumes of text-based tasks with high accuracy and consistency.

2024 - 2024

Search Relevance & Safety Annotation – DataAnnotationTech

Data Annotation TechTextEntity Ner ClassificationClassification

Judged the relevance and correctness of search results, classified responses based on intent match, accuracy, and contextual quality. Annotated sensitive content and applied safety guidelines to identify policy violations. Evaluated query–result alignment and flagged misleading, harmful, or low-quality outputs. Maintained consistency across large datasets while adhering to strict quality and speed metrics.

Judged the relevance and correctness of search results, classified responses based on intent match, accuracy, and contextual quality. Annotated sensitive content and applied safety guidelines to identify policy violations. Evaluated query–result alignment and flagged misleading, harmful, or low-quality outputs. Maintained consistency across large datasets while adhering to strict quality and speed metrics.

2023 - 2023

OCR & Machine Learning Research Project

OtherImageClassification

Labeled OCR output errors and classified recurring misrecognition patterns. Evaluated model predictions, refined training dataset structure, and improved dictionary learning performance by 15.42%. Documented annotation logic, maintained dataset versions, and contributed to analysis of algorithmic failure cases. Accelerated iteration cycles and improved overall recognition accuracy.

Labeled OCR output errors and classified recurring misrecognition patterns. Evaluated model predictions, refined training dataset structure, and improved dictionary learning performance by 15.42%. Documented annotation logic, maintained dataset versions, and contributed to analysis of algorithmic failure cases. Accelerated iteration cycles and improved overall recognition accuracy.

2022 - 2023

Show more

Education

Work History