Riko K.

Key Skills

Software

SuperAnnotate

Top Subject Matter

LLM evaluation

RLHF auditing

Supervised fine-tuning datasets

Top Data Types

Text

Video

Document

Top Task Types

RLHF

Fine-tuning

Red Teaming

Evaluation/Rating

Freelancer Overview

Bilingual professional with over 5 years of cross-functional experience bridging technical support, digital media, and high-precision AI operations. Specializing in Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT) auditing, this expertise encompasses rigorous data quality assurance, prompt engineering, and language model evaluation. Proven track record in analyzing complex LLM outputs, eliminating hallucinations, enforcing strict instruction-following constraints, and conducting adversarial red-teaming to maximize model safety. A unique combination of technical precision and advanced communication skills sets this profile apart. Fluent in both English and Japanese, bringing a sharp cross-cultural perspective to data curation, localization, and multilingual evaluation tasks.

Labeling Experience

RLHF Reviewer / SFT Auditor (March 2026 - Present)

MercorTextRLHF

Evaluated LLM responses by applying precise data labeling to improve output quality and reliability. Used RLHF-based ranking to assess helpfulness, truthfulness, safety, and adherence to constraints. Authored reference responses and golden datasets to support supervised fine-tuning and advanced domain reasoning. • Ranked model outputs using RLHF methodologies for evaluation and grading • Performed fact-checking and hallucination detection to validate responses • Conducted quality assurance checks on tone, accuracy, and logical consistency • Produced labeled reference materials (golden datasets) for training support

2026 - Present

AI Language Consultant (August 2025 - October 2025)

Text

Provided bilingual linguistic review and quality feedback on AI model outputs in Japanese and English. Focused on correctness and clarity through structured review workflows under quality standards. Supported improvement of AI output quality by supplying targeted review notes. • Reviewed AI-generated text for Japanese–English linguistic quality • Assessed accuracy and clarity of model outputs through bilingual feedback • Logged and communicated quality issues to improve future responses • Worked within structured review workflows and documentation expectations

2025 - 2025

Education

Work History