At Mercor, I train and validate AI models across data domains, using prompt tuning, consistency checks, and error analysis to improve classification accuracy. I also developed and validated safety prompt annotation guidelines that achieved 92% compliance in spot audits.
At Outlier, I evaluate LLM responses for accuracy, reasoning, instruction following, relevance, and localization. I provide actionable feedback on errors and recurring patterns to support better model training.
Through Upwork, I’ve translated, proofread, interpreted, and transcribed materials in Indonesian, English, and Korean across legal, medical, technical, cultural, and marketing topics. I’ve also annotated transcription data at Chemin AI and worked as a translator and interpreter across other projects and organizations.
