At Various AI Research Platforms, I’ve completed 2,500+ evaluations of generative AI outputs across mathematics, coding, data analysis, and technical reasoning. I also verified solutions step by step and ran 340+ paired model comparisons.
I authored a 47-point evaluation rubric adopted by a 12-person evaluator team, raising inter-rater agreement from 72% to 89%. I produced RLHF preference-labeling data and designed red-teaming exercises that surfaced 23 previously unidentified failure modes.
I built Python tools to automate evaluation workflows, cutting manual processing time by about 35%. I also wrote evaluation reports that informed development priorities and training-data curation.
As a freelance Data Analyst and Digital Content Strategist, I delivered 28 data analysis projects and developed a PySpark/MLlib customer segmentation model that lifted campaign conversion by 18%. My background also includes census data quality and QA work with Kenya National Bureau of Statistics (KNBS).
