AI Trainer (Part-Time) at Invisible Technologies
Text
Prompt + Response Writing (SFT)Generated high-quality prompts and ideal responses to support supervised fine-tuning (SFT) and improve large language model (LLM) response quality and reasoning capabilities. Conducted RLHF-style evaluation by ranking and comparing model outputs against detailed rubrics to select the most accurate and aligned responses. Annotated datasets and produced structured training data to improve model accuracy, safety, and instruction-following behavior. • Created and standardized rubrics and evaluation criteria for consistent scoring across training tasks • Performed prompt testing and model behavior analysis to identify failure modes like hallucinations and instruction misalignment • Built structured training examples intended for SFT and alignment • Coordinated with distributed teams for iterative training and continuous improvement cycles