Kang X.

Key Skills

Software

Label Studio

Labelbox

CVAT

Top Subject Matter

Insurance Domain Expertise

Financial Large Language Models (LLMs)

Legal Documents

Top Data Types

Text

Document

Image

Top Task Types

RLHF

Entity (NER) Classification

Classification

Polygon

Freelancer Overview

LLM RLHF & SFT Data Labeling Specialist. Brings 7+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Internal and Proprietary Tooling. Education includes Master of Science, The Chinese University of Hong Kong and Bachelor of Engineering / Economics, Zhejiang University. AI-training focus includes data types such as Text, Document, and Image and labeling workflows including RLHF, Entity (NER) Classification, and Classification.

Labeling Experience

Annotation Automation & Quality Assurance Lead

Developed automated annotation workflows leveraging AI pre-labeling, with human experts focused on auditing and correction. Introduced robust three-tiered QA procedures and implemented Gold Set (Blind Test) evaluations to ensure data quality. Boosted annotation throughput and maintained zero-pollution standards for training datasets. • Incorporated pre-labeling via existing models. • Implemented strict multi-stage QA system. • Used blind test sets for real-time monitoring. • Tripled annotation productivity and data cleanliness.

Present

Claims Data Annotation & Active Sampling Specialist

ImagePolygon

Managed annotation projects for multi-modal insurance claims, focusing on auto salvage parts and medical invoice imagery. Established precise segmentation baselines through polygon annotation and structured OCR alignment for automated claim assessment. Applied hard sample mining strategies to increase annotation value and efficiency. • Polygon segmented damage in auto parts images. • Structured OCR text from medical bills. • Hard sample selection via active learning scripts. • Enhanced generalization by prioritizing challenging samples.

Present

Intent Classification Annotation Specialist

TextClassification

Constructed a semantic intent classification hierarchy using real customer dialogue samples from a leading insurance company. Annotated and classified intents across 38 sub-categories for advanced dialogue system development. Supported intelligent traffic routing and precise recognition for AI-driven customer support. • Based on real-world insurance inquiries. • Refined deep semantic intent structures. • Enhanced system accuracy and traffic handling. • Supported AI-enabled customer workflows.

Present

Named Entity Recognition (NER) Annotation Lead

DocumentEntity (NER) Classification

Designed a fine-grained Named Entity Recognition (NER) annotation framework for unstructured insurance contracts and legal documents. Directed the extraction of critical fields such as Coverage Scope, Deductibles, and Payout Ratios to increase labeling accuracy. Set new accuracy benchmarks for ground truth extraction in the insurance domain. • Focused on legal and insurance document NER. • Improved coverage of complex contract clauses. • Ensured compliance with domain-specific regulations. • Delivered industry-leading ground truth data quality.

Present

LLM RLHF & SFT Data Labeling Specialist

TextRLHF

Designed and executed RLHF ranking and reward model training tasks for insurance and financial LLMs. Developed high-quality comparison datasets by implementing advanced, multi-dimensional scoring systems. Ensured model safety and compliance through adversarial annotation and output alignment with regulatory requirements. • Created datasets targeting Factuality, Compliance, and Helpfulness. • Led instruction-tuning with over 20,000 vertical domain samples. • Enforced precise safety annotation and red teaming for risk. • Enhanced model professional reasoning with expert supervision.

Present

Education

Work History