This includes engaging with substantial and varied datasets—such as CSVs, PDFs, spreadsheets, and technical documentation—to pose rigorous, multi-step problems that challenge how AI interprets, reasons, and acts in genuine market scenarios. Develop realistic evaluation tasks for AI agents, simulating complex activities in securities and commodities trading, client advisory, and compliance reasoning.