AI Annotation Strategy
My approach encompasses rubric construction and evaluation, fine-tuning Golden Responses and casting a critical human eye over responses to verify honesty and integrity.
Read More→
My work sits at the intersection of language, instruction and judgement. By training models through RLHF (Rehearsed Learning through Human Feedback), my focus is to ensure that every LLM response produced checks the "4U" criteria - unequivocally correct, uniform, useful and understandable.
My approach encompasses rubric construction and evaluation, fine-tuning Golden Responses and casting a critical human eye over responses to verify honesty and integrity.
Read More→Studying the edge cases where helpfulness, honesty and safety collide, and producing reference responses that resolve them without losing warmth.
Read More→Accessible language to teach models that learn. Multi-step reasoning, gentle correction and the metaphors that make difficult ideas hold in the mind.
Read More→Verifying that every response meets the standard. The feedback loops, comparison frameworks and the attention to detail that keeps annotation honest at scale.
Read More→