Agricultural AI Training Scenarios (Mercor)

This sample is the most directly relevant to Contra Labs. During my contract with Mercor as a Soil and Plant Scientist, I developed structured real-world agricultural scenarios to train and evaluate frontier AI models on soil science, crop physiology, and experimental design reasoning. I also reviewed and validated AI-generated outputs for scientific accuracy and field applicability — catching hallucinated statistics, misclassified soil types, and agronomically incorrect recommendations. This experience means I already understand what good AI training data looks like, what failure modes to watch for, and how to write precise, expert-level feedback that improves model quality.