Richard Bonar

Key Skills

Software

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Text Generation

RLHF

Evaluation Rating

Prompt Response Writing SFT

Freelancer Overview

I am an experienced evaluator, scenario writer, and AI trainer with a strong academic background in applied mathematics, physics, and computer science. My work focuses on designing and reviewing high-quality training data for AI and LLM systems, including writing clear prompts, step-by-step solutions, and rubrics for math and physics domains. I have hands-on experience with leading data labeling and annotation platforms such as Remotasks, Scale AI, Toloka, and more, and I am adept at using structured formats like JSON and YAML for scenario specification and QA. My technical skills include Python, JavaScript, and version control with Git, which I use to validate, parse, and support data quality. I excel at identifying edge cases, ensuring rigorous evaluation, and improving prompt and scoring guidelines based on observed model performance and failure modes.

Labeling Experience

Freelance AI Trainer / LLM Evaluator (Math and Physics)

Scale AITextText GenerationRLHF

Rubric-based evaluation and training data creation for math and physics LLM tasks. Rated model/agent outputs for correctness, completeness, instruction-following, safety, and clarity, and documented scoring rationale. Wrote and audited step-by-step solutions with correct notation and unit/dimensional checks. Produced corrected answers with concise explanations, tagged items with topic/difficulty/reasoning/error metadata, and flagged edge cases or ambiguous items to improve scoring reliability.

Rubric-based evaluation and training data creation for math and physics LLM tasks. Rated model/agent outputs for correctness, completeness, instruction-following, safety, and clarity, and documented scoring rationale. Wrote and audited step-by-step solutions with correct notation and unit/dimensional checks. Produced corrected answers with concise explanations, tagged items with topic/difficulty/reasoning/error metadata, and flagged edge cases or ambiguous items to improve scoring reliability.

2023 - 2025

Education

Work History