At Dropbox Dash, I rate AI responses and enterprise search results—including documents, images, and video—for relevance, accuracy, and factual grounding. I check answers against internal documents and employee communications to verify their claims.
When ambiguous queries or edge cases arise, I apply and update the evaluation rubric so labels stay consistent across evaluators. I also use Dash Assistant, internal search tools, and KQL-style queries to trace sources and diagnose relevance problems.
At Google Gemini, through GlobalLogic, I evaluated text, image, audio, and video outputs against user intent and program rubrics. I provided human labels for prompt engineering and model tuning, and contributed to evaluator training and documentation.
Earlier, I designed lessons, scripts, and assessments at VIPKid, translating complex subject matter into accessible content. My work has also included rubric-based feedback as an English Instructor at YBM ECC.
