At Handshake AI Fellowship, I develop multimodal prompts across three projects to evaluate and train LLMs and agentic AI models. I design challenging question-and-answer pairs to assess visual reasoning and real-world knowledge.
I also create realistic tasks in persona universes, compare model outputs, and score performance with a structured rubric. My conversational prompts test specific skills and assess communication and accuracy between two models.
At Micro1, I annotate robotic-task videos to help train models to recognize successful task execution and errors. At SMECareers, I audit English transcriptions, identify errors, and provide structured feedback using guidelines for normalization, disfluencies, and paralinguistic tokens.
Earlier, at the UCDavis PlantSciences Department's Schmitt Lab, I examined herbarium specimens and labeled plant data for machine-learning training. I also investigated climate change and apical tuft growth, analyzed data, and presented my findings.
