Kd User

I am a Maths Subject-Matter Expert with hands-on experience training and evaluating AI systems in mathematical reasoning. I design exam-style MCQs, develop clear annotation rubrics, and produce high-quality training data for NLP and math-evaluation pipelines.

At my current role I designed and validated over 3,000 math items including worked solutions and stepwise answer keys, and authored annotation guidelines that reduced labeler disputes by 42%. I perform error analysis on model outputs and create targeted datasets, improving model accuracy on benchmark tasks by approximately 6–8%.

Previously, I created curriculum-linked practice sets and custom worksheets for grades 4–12 and competitive prelims, drove average student score improvements of 12–18% in two months, and delivered live doubt-clearing sessions and explainer videos. I also curated an 800+ example edge-case pack to improve annotator consistency and model robustness.

I collaborate closely with engineers to define evaluation metrics and inputs for model releases, and I maintain a reusable rubric library for multi-step proofs and numerical reasoning. I seek roles where I can contribute domain expertise, annotation governance, and measurement-driven improvements to AI math reasoning pipelines.