Yi W.

Key Skills

Software

Internal/Proprietary Tooling

Top Subject Matter

Model output evaluation

Preference ranking

Data annotation

Top Data Types

Video

Image

Top Task Types

Classification

Segmentation

Bounding Box

Freelancer Overview

In terms of constructing and evaluating high-quality AI training data, I have practical experience from underlying data engineering to quality control of generative model output. In my current project on controllable generation and lesion recognition of fundus images, I independently led a rigorous data preprocessing chain. Addressing the challenging shortcomings of raw medical data, such as large scale variance and incomplete field of view, I innovatively applied contact chord length detection and proportional filling strategies to clean and standardize more than 3,000 noisy raw data images into a high-quality, morphologically distortion-free training set of 512x512.

Education

Work History