At Mercor, I designed architecture-domain adversarial prompts to test AI model performance against graded compliance criteria. I developed a residential compliance-review task and refined scoring rubrics and golden-answer references against ground truth.
At Genius Hub Global Initiative, I manage and profile a 600+ participant dataset, building Excel tracking tools and preparing outputs for programme reporting.
I've also evaluated LLM text and code, ranked preferences for RLHF, and annotated AI-generated images and real-world photos. Earlier, I produced technical drawings and 3D models during architecture internships using AutoCAD and Autodesk Revit.

