Hai Z.

Key Skills

Software

Other

Top Subject Matter

E-commerce product reviews

user preferences

RLHF-oriented output quality evaluation

Top Data Types

Text

Image

Top Task Types

Fine-tuning

RLHF

Data Collection

Freelancer Overview

VOCReview AnnotationSystem — Senior Algorithm Engineer (Data Intelligence Center). Brings 9+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal, Proprietary Tooling, and Other. Education includes Master of Science, Texas A&M University (2018) and Bachelor of Engineering, Huazhong Agricultural University (2015). AI-training focus includes data types such as Text, Image, and Medical and labeling workflows including Evaluation, Rating, and Fine-tuning.

Labeling Experience

AI Output Evaluation & Prompt Engineering — Senior Algorithm Engineer

OtherTextRLHF

Performed AI output evaluation for LLM-generated content including relevance, factual accuracy, tone, and business appropriateness. Iterated on prompt strategies and conducted large-scale output ranking and benchmarking across production LLM tasks. Focused the feedback and labels on quality and safety signals suitable for RLHF-style reward modeling. • LLM output scoring and comparative evaluation (helpfulness, accuracy, coherence/tone, safety) • Prompt refinement and output quality evaluation across production tasks • Internal prompt benchmarking framework to compare model versions • Feedback collection aligned to RLHF evaluation needs

2022 - Present

Multi-Modal Image Attribute Annotation Pipeline — Senior Algorithm Engineer

OtherImageFine-tuning

Built a large-scale image attribute labeling pipeline to annotate millions of fashion product images across 16+ visual attributes. Implemented Qwen3-VL-based labeling with normalization and conflict-resolution logic to reduce ambiguity at scale. Used the annotated data as training/evaluation input for CLIP/SigLIP contrastive learning fine-tuning to improve retrieval ranking quality. • Multi-dimensional fashion image attribute annotation (16+ attributes) • Qwen3-VL-PluS-based automated labeling pipeline • Label normalization and conflict resolution / QC workflows • Training/evaluation support for CLIP/SigLIP contrastive learning fine-tuning

2022 - Present

VOCReview AnnotationSystem — Senior Algorithm Engineer (Data Intelligence Center)

Text

Designed and executed LLM-assisted evaluation of product user reviews using a dual-axis framework covering product attribute sentiment tags and user persona tags. Built structured datasets from the resulting annotations to support downstream personalization and recommendation training. Ensured label usability by defining annotation guidelines, data schemas, and iteratively refining label definitions with business stakeholders. • Dual-axis review annotation (positive/negative product attribute evaluation + user portrait/persona tags) • Human-in-the-loop validation for multi-dimensional labeling accuracy • Annotation schema design and inter-iteration refinement • Produced structured preference datasets for R&D insights and model training

2022 - Present

Medical NER Dataset Annotation — NLP Algorithm Engineer

Supervised and participated in building medical NER annotation datasets by defining entity labeling schemas for diseases, drugs, symptoms, and hospitals. Coordinated annotation teams and validated inter-annotator consistency to maintain high-quality ground truth. The labeled corpus supported end-to-end training and validation of a BERT + BiLSTM + CRF model achieving strong entity extraction performance. • Medical NER schema definition for biomedical entities • Team coordination and inter-annotator agreement validation • Dataset construction for model training and evaluation • Annotation-to-model impact via >95% entity extraction F1

2019 - 2021

Graduate Research Associate — Bioscience & Technology

TextData Collection

Contributed to biomedical knowledge graph annotation and structured data labeling in collaboration with MD Anderson Cancer Center. Labeled scientific biomedical information such as cancer signaling pathways and drug-receptor interactions with emphasis on precision and documentation quality. Produced structured annotations suitable for downstream knowledge representation and AI research workflows. • Biomedical knowledge graph annotation • Structured labeling for cancer signaling pathways and drug-receptor interactions • Academic-level annotation precision and documentation standards • Collaboration with medical research partners to ensure dataset quality

2015 - 2018

Education

Work History