Yikuai B.

Key Skills

Software

AWS SageMaker

Axiom AI

Clickworker

Top Subject Matter

LLM Output Evaluation

Rlhf Domain Expertise

Prompt Quality Annotation

Top Data Types

Text

Document

Top Task Types

Classification

RLHF

Question Answering

Text Generation

Freelancer Overview

Independent AI Engineer (LLM Output Evaluation & Annotation). Brings 2+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Internal and Proprietary Tooling. Education includes Bachelor of Science, 哈尔滨商业大学 (2027). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Classification.

Labeling Experience

Independent AI Engineer (LLM Output Evaluation & Annotation)

Text

As an independent AI Engineer, I evaluated outputs from multiple Large Language Models (LLMs), including Claude, Qwen, and GPT-4, to improve performance. I developed internal rubrics and methodologies for judging adherence to instructions, factual accuracy, tone consistency, and hallucination detection. My work involved comparative assessments and prompt-based iterations to systematically enhance model responses. • Conducted bilingual (Mandarin/English) annotation and evaluation for RLHF and LLM output comparison tasks. • Created and applied original evaluation rubrics for LLM preference data and error analysis. • Executed prompt-quality annotation and side-by-side model judgment across multiple scenarios and user queries. • Documented prompt engineering process and published technical guides for the developer community.

2024 - Present

Data Annotator & Classifier Trainer (TripAgent LLM System)

TextClassification

In the TripAgent project, I manually labeled over 500 Xiaohongshu UGC posts using a quality tier annotation schema. This schema categorized posts into high-signal, promotional, and low-quality classes intended to inform filtering and preference data tasks. The resulting labeled dataset contributed to training a Chinese RoBERTa classifier and establishing benchmarks for LLM output evaluation. • Designed a content quality annotation schema directly analogous to LLM preference labeling. • Hand-labeled and validated data used for training a supervised classification model. • Enabled automation of Chinese-language content filtering for downstream AI experiments. • Built evaluation harnesses for comparative LLM generation analysis using this annotated data.

2025 - 2025

Education

Work History