Freelancer

Key Skills

Software

No software listed

Top Subject Matter

Radiology

Top Data Types

Image

Text

Medical Dicom

Top Task Types

Prompt + Response Writing (SFT)

Evaluation/Rating

RLHF

Diagnosis

Classification

Question Answering

Freelancer Overview

I work as a medical subject-matter expert on AI model training and evaluation, contributing to clinical and general large-language-model projects through expert networks including Mercor, Outlier, and Handshake AI. The core of the work is human preference labeling: for each prompt, I assess competing model responses across six structured dimensions (instruction following, truthfulness, correctness, verbosity, writing quality, and overall quality), score each on a defined scale, then select a preferred response and write an evidence-based justification grounded in specific clinical and technical detail. Drawing on my diagnostic radiology and physician training, I focus on catching factual and reasoning errors that fluent, well-formatted answers can mask, applying the principle that clinical correctness outweighs polish. This work spans a wide range of clinical domains, including case interpretation, hepatology, palliative and goals-of-care consults, emergency and trauma scenarios, pharmacy workflows, and imaging, as well as non-clinical areas such as operations, legal analysis, nutrition, and information security. Beyond scoring, I author "golden" reference responses, help build and refine evaluation rubrics, and identify designed edge cases and failure modes (for example, confidentiality pitfalls and internal inconsistencies in clinical documentation) that separate high-quality from low quality model output. Together this combines hands-on clinical and medical-imaging expertise with practical familiarity with RLHF-style human-feedback pipelines, annotation standards, and model-evaluation methodology.

Education

Work History