Ulam builds research-grade reasoning datasets and evaluation systems for frontier AI labs and AI developers. Its work includes human-in-the-loop reasoning traces, expert-reviewed proof attempts, and private benchmark programs.