Peter M.

Key Skills

Software

Clickworker

Appen

Don't disclose

Top Subject Matter

Legal contracts

financial statements

Accounting/invoices Domain Expertise

Top Data Types

Image

Audio

Geospatial Tiled Imagery

Document

Text

Top Task Types

Question Answering

Object Detection

Text Summarization

Transcription

Data Collection

Function Calling

RLHF

Freelancer Overview

Document Annotation & Data Extraction Specialist (AI training support via RLHF, schema validation). Core strengths include Don't disclose. Education includes Bachelor of Science, N/A. AI-training focus includes data types such as Document and Text and labeling workflows including RLHF, Translation, and Localization.

Labeling Experience

Document Annotation & Data Extraction Specialist (AI training support via RLHF, schema validation)

Don't discloseDocumentRLHF

Performed document data extraction and annotation for legal, financial, accounting, and compliance materials from PDF and DOCX sources. Validated AI-generated extraction outputs against the source documents to ensure alignment with JSON schemas and corrected inconsistencies, missing fields, and formatting errors. Applied RLHF methodologies to improve model response quality and support AI training workflows. • Extracted structured entities such as dates, clauses, financial values, signatures, and account information • Processed dynamic and repeated entries like transaction line items and payment schedules • Ensured correct JSON schema handling including arrays, nested objects, and null values • Conducted quality assurance checks for consistency and reduced extraction errors

2023 - 2025

Generative AI prompt engineering, hallucination checking, and rubric-based evaluation for LLM outputs

Don't discloseText

I performed AI output evaluation by identifying logical errors and hallucinations in model-generated text, then revising responses to satisfy project-specific quality and safety expectations. I ranked and scored outputs against strict rubrics covering logic, tone, and safety requirements. I ensured the final responses met the “Helpful, Honest, and Harmless” standard expected in high-level AI training workflows. • Detected hallucinations and reasoning mistakes in AI-generated text • Revised outputs to improve accuracy, nuance, and guideline compliance • Evaluated responses against rubrics for logic, tone, and safety • Maintained consistency with model style guides and project instructions

2022 - 2024

Education

Work History