Havon M.

Key Skills

Software

Remotasks

Top Subject Matter

Forensic sciences and quantitative/academic domains (toxicology, chemistry, DNA analysis, biometrics, ballistics)

Advanced scientific

Quantitative Domain Expertise

Top Data Types

Text

Top Task Types

Prompt + Response Writing (SFT)

Freelancer Overview

AI Specialist / Data Annotator (LLM evaluator) — Outlier AI & Remotasks. Core strengths include Outlier AI, Remotasks, and Handshake AI. Education includes Bachelor of Science, N/A and High School Diploma, N/A. AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Prompt + Response Writing (SFT).

Labeling Experience

AI Content Evaluator — Handshake AI

TextPrompt + Response Writing (SFT)

Reviewed and optimized machine-generated content by rewriting responses to achieve high standards of clarity, tone, and technical precision. Served as a subject matter expert to assess model performance and accuracy for advanced scientific, quantitative, and technical inquiries. Contributed to dataset creation by delivering cleaned outputs within asynchronous, high-priority training pipelines. • Rewrote model responses for improved readability and technical correctness. • Evaluated accuracy and performance for SME-level scientific/quantitative questions. • Collaborated asynchronously to maintain dataset quality for training cycles. • Ensured outputs were structurally sound and aligned with domain expectations.

Present

AI Specialist / Data Annotator (LLM evaluator) — Outlier AI & Remotasks

Text

Conducted LLM evaluation and annotation by grading model responses to complex, multi-turn prompts against truthfulness, helpfulness, and safety criteria. Applied Likert-scale psychometric frameworks to rate factuality, logical coherence, and adherence to strict prompt requirements. Produced corrected outputs by identifying subtle hallucinations and logical/structural errors for downstream training and optimization. • Reviewed multi-turn prompt/response pairs and scored outputs. • Used Likert scales and constraint/adherence checks for consistency. • Drafted adversarial prompt sets to test edge-case behavior in scientific and quantitative domains. • Categorized and corrected hallucinations, logical fallacies, and formatting issues.

Present

Education