王鑫宇

Key Skills

Software

Label Studio

CVAT

OpenCV AI Kit (OAK)

LabelImg

Top Subject Matter

AI Model Evaluation

Programming Code

Dialogue RLHF

Top Data Types

Text

Document

Image

Top Task Types

RLHF

Classification

Freelancer Overview

AI Data Annotation Specialist. Core strengths include Label Studio, Internal, and Proprietary Tooling. Education includes Bachelor of Science, San Jose State University (2023). AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including RLHF, Evaluation, and Rating.

Labeling Experience

AI Data Annotation Specialist

Label StudioRLHF

As an AI Data Annotation Specialist at Alegion, I performed Reinforcement Learning from Human Feedback (RLHF) tasks on AI-generated code and dialogues. My work included ranking model outputs, writing demonstration dialogues, and verifying logical reasoning of large language models. I consistently delivered high-quality annotations using industry-standard tools and internal platforms. • Evaluated and debugged AI-generated Python and Java code for syntax, efficiency, and security. • Conducted Chain-of-Thought verifications to assess model reasoning and logic. • Red-teamed edge cases to assess model robustness and accuracy. • Processed over 5,000 prompts and responses, achieving an audited annotation accuracy of 98.5%.

2024 - Present

LLM Agentic Workflow Evaluation & Integration (Project)

Text

During the LLM Agentic Workflow Evaluation & Integration project, I evaluated multiple AI APIs and annotated generated responses. I conducted manual A/B testing to determine quality, coherence, and instruction-following capabilities of different large language models. My analysis helped identify strengths and weaknesses in model outputs connected to RAG pipelines. • Compared outputs from OpenAI, Gemini, Anthropic, and DeepSeek APIs. • Assessed factual grounding and contextualization from vector database inputs. • Rated instruction-following and coherence of model responses. • Delivered annotated findings to guide further model selection and integration.

2025 - 2026

Financial Time-Series & Computer Vision Data Processing (Project)

CVATDocumentClassification

In the Financial Time-Series & Computer Vision Data Processing project, I processed quantitative financial data and annotated visual datasets for training and evaluation. My tasks included labeling trend reversals, anomalies, and providing bounding box annotations using the YOLO object detection framework. These efforts supported model development for trading algorithms and vision models. • Labeled financial time-series data for backtesting strategies. • Drew and refined bounding boxes and polygons for object detection. • Researched and implemented advanced annotation methodologies. • Ensured high precision and accuracy through quality control processes.

2024 - 2025

Education

Work History