Current Projects
Digital Twins, Agentic AI, and Limit Testing
Testing the Twins - 2k500 Twin Data Set
A persona-modeling study that uses the Twin-2K-500 dataset to build language-model personas and test how they hold up under controlled, near-identical conditions. The work probes the limits of synthetic respondents in data-limited environments, asking where model-driven personas stay reliable and where they break.
ATLAS - Mapping the Machine Workforce
A system that helps corporations adopt agentic AI by first translating what agents can and cannot do for their business. It maps an organization into its full task structure across finance, operations, people, technology, marketing, and sales, then overlays jobs and agents to show which tasks are ready to hand off and which are not. The map is designed and scaled to each business. From there, it guides where to build multi-agent systems that take on real work. The Agent Atlas turns agentic adoption from a guess into a clear, navigable picture. The Business Task Atlas, shown here, is the map at its core.A system that helps corporations adopt agentic AI by first translating what agents can and cannot do for their business. It maps an organization into its full task structure across finance, operations, people, technology, marketing, and sales, then overlays jobs and agents to show which tasks are ready to hand off and which are not. The map is designed and scaled to each business. From there, it guides where to build multi-agent systems that take on real work. The Agent Atlas turns agentic adoption from a guess into a clear, navigable picture. The Business Task Atlas, shown here, is the map at its core.
Please email me for access.
Synthetic Worldbuilding - Politics, Policy, and Strategy
With the Carnegie Mellon Institute for Strategy & Technology and OFAI at Carnegie Mellon University. A system that builds synthetic worlds from agents, synthetic data, and digital twins, then runs them forward to train decision-makers. By generating many outcomes and contingencies from a single decision point, it lets executives, military leaders, and policymakers rehearse choices and see how they play out before they act.
Technology and Engineering Strategy
Finding Life - A K-Means Approach
An unsupervised learning study using K-Means clustering to uncover latent structure in high-dimensional habitability and life-detection data. The project focuses on exploratory pattern discovery where labels are unavailable or uncertain, demonstrating how clustering can surface meaningful regimes without prior assumptions.
Technology and Engineering Entrepreneurship and Innovation
BaseNLPSP500
A baseline NLP framework examining whether textual sentiment contains predictive signal for S&P 500 movements. The project emphasizes transparent preprocessing, sentiment scoring, and defensible modeling as a foundation for more advanced market-facing NLP systems.