Adrian Chambers

Key Skills

Software

Data Annotation Tech

Google Cloud Vertex AI

Labelbox

Scale AI

Snorkel AI

Other

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Text

Top Label Types

Computer Programming Coding

Data Collection

Evaluation Rating

Fine Tuning

Function Calling

Prompt Response Writing SFT

Question Answering

Red Teaming

RLHF

Text Generation

Text Summarization

Freelancer Overview

I specialize in AI/ML evaluation, data labeling, and training data quality, with hands-on experience designing and executing golden datasets for LLM benchmarking and fine-tuning. My background includes reviewing and annotating AI-generated SQL queries, performing systematic quality assessments of frontier AI models, and developing detailed evaluation rubrics for accuracy, completeness, and logical flow. I have led red teaming and adversarial testing campaigns to identify model vulnerabilities and documented diverse AI failure modes, ensuring robust and reliable model performance. My technical skills span Python, SQL, data validation, and quality control tools such as pandas, pytest, and Git, enabling me to deliver high-confidence training data and annotation for NLP and code generation domains. I am committed to maintaining rigorous standards in data annotation and AI training workflows to support the development of safe and effective AI systems.

Labeling Experience

Software Engineer AI Trainer

OtherComputer Code ProgrammingEvaluation RatingComputer Programming Coding

Analyzed and documented open-source pull requests to create high-quality training data for AI coding agents, ensuring comprehensive understanding of code changes and their implications Designed and authored precise issue descriptions that enable AI agents to reproduce software fixes, balancing behavioral specifications with implementation details to achieve optimal pass rates Developed and validated Docker-based test environments for reproducible AI agent evaluation, including dependency management, build configuration, and test execution infrastructure Created oracle test suites by extracting and adapting tests from PRs, ensuring proper isolation and compatibility with AI evaluation pipelines Iteratively refined issue descriptions based on AI agent performance analysis, identifying and resolving false negatives (missing requirements) and false positives (untested specifications) Reviewed and annotated AI agent attempts to verify solution correctness, distinguishing between genuine age

Analyzed and documented open-source pull requests to create high-quality training data for AI coding agents, ensuring comprehensive understanding of code changes and their implications Designed and authored precise issue descriptions that enable AI agents to reproduce software fixes, balancing behavioral specifications with implementation details to achieve optimal pass rates Developed and validated Docker-based test environments for reproducible AI agent evaluation, including dependency management, build configuration, and test execution infrastructure Created oracle test suites by extracting and adapting tests from PRs, ensuring proper isolation and compatibility with AI evaluation pipelines Iteratively refined issue descriptions based on AI agent performance analysis, identifying and resolving false negatives (missing requirements) and false positives (untested specifications) Reviewed and annotated AI agent attempts to verify solution correctness, distinguishing between genuine age

2025

Web Development / Cybersecurity Subject Matter Expert (SME)

OtherComputer Code ProgrammingComputer Programming Coding

Created AI grader rubrics, system instructions, and test cases for Coursera's Web Development and Cybersecurity courses.

Created AI grader rubrics, system instructions, and test cases for Coursera's Web Development and Cybersecurity courses.

2025

Software Engineer AI Trainer

Snorkel AIComputer Code ProgrammingEvaluation RatingComputer Programming Coding

Evaluated AI model outputs on real-world code changes across open-source repositories. As Submitter, crafted multi-turn prompts and rated model trajectories across multiple quality dimensions with evidence-based justifications. As SWE Reviewer, validated automated evaluations, verified claims against code diffs, and made accept/reject decisions. As Adjudicator (Final Reviewer), served as the final decision-maker in a three-tier review pipeline, synthesizing multiple reviewer assessments, overturning incorrect decisions when evidence warranted, and providing actionable feedback to reviewers.

Evaluated AI model outputs on real-world code changes across open-source repositories. As Submitter, crafted multi-turn prompts and rated model trajectories across multiple quality dimensions with evidence-based justifications. As SWE Reviewer, validated automated evaluations, verified claims against code diffs, and made accept/reject decisions. As Adjudicator (Final Reviewer), served as the final decision-maker in a three-tier review pipeline, synthesizing multiple reviewer assessments, overturning incorrect decisions when evidence warranted, and providing actionable feedback to reviewers.

2025

Software Engineering / Data Science AI Trainer

Scale AIComputer Code ProgrammingFine TuningEvaluation Rating

Designed adversarial data analysis prompts across 18+ domain areas to identify AI model failure modes Created comprehensive evaluation rubrics with necessary and sufficient success criteria Developed ground truth solutions with reproducible code in Google Colab for validation Identified and documented 10 major failure categories including calculation errors (40%), data handling issues (25%), and methodological errors (20%) Generated complex statistical visualizations including subplots, distributions, and multi-dimensional analyses Performed advanced statistical analyses including regression (70% of tasks), hypothesis testing (80%), classification (60%), and time series analysis (30%) Documented quality control procedures covering unit conversions, data alignment, and temporal analysis Identified common AI failure patterns including cascading errors, magnitude mistakes (orders of magnitude off), and method substitution issues

Designed adversarial data analysis prompts across 18+ domain areas to identify AI model failure modes Created comprehensive evaluation rubrics with necessary and sufficient success criteria Developed ground truth solutions with reproducible code in Google Colab for validation Identified and documented 10 major failure categories including calculation errors (40%), data handling issues (25%), and methodological errors (20%) Generated complex statistical visualizations including subplots, distributions, and multi-dimensional analyses Performed advanced statistical analyses including regression (70% of tasks), hypothesis testing (80%), classification (60%), and time series analysis (30%) Documented quality control procedures covering unit conversions, data alignment, and temporal analysis Identified common AI failure patterns including cascading errors, magnitude mistakes (orders of magnitude off), and method substitution issues

2025 - 2025

Software Engineer / Cybersecurity AI Trainer

Data Annotation TechComputer Code ProgrammingEvaluation RatingRed Teaming

Worked on over 15+ projects on the platform in the domains of computer science, software engineering, data science, cybersecurity, and educational technology. Projects included side-by-side comparisons, code review/editing, model red teaming, evaluation rubric creation, quality assessments, and more.

Worked on over 15+ projects on the platform in the domains of computer science, software engineering, data science, cybersecurity, and educational technology. Projects included side-by-side comparisons, code review/editing, model red teaming, evaluation rubric creation, quality assessments, and more.

2023 - 2025

Education

Work History