Hello! My name is Anthony and I am an AI researcher focused on using AI for high-impact real-world problems. Recently, I have been developing decision-support tools for the development of large-scale sustainable resource projects such as mineral exploration, geothermal production, and carbon sequestration, where AI can improve both efficiency and safety. As part of an interdisciplinary team, we model these problems as large partially observable Markov decision processes (POMDPs) and develop custom solvers for them. Previously, I have worked on the validation of safety critical autonomy in domains such as autonomous driving and aviation.

In my research I focus on three key pillars that enable the development of reliable autonomy for high-stakes applications

1. Building Robust Models: Modeling the environment is an essential step toward developing AI systems that can effectively operate in it. For example, in our recent paper we developed a deep neural network surorogate model of CO₂ reservoir dynamics to plan for geological carbon storage.
2. Planning Under Uncertainty: With the models developed, we need algorithms to plan optimal actions. See our recent work BetaZero, which combines planning and learning to solve POMDPs.
3. Extensive Safety Validation: Once the AI system has been developed, it needs extensive testing before deployment. See this survey on algorithms for automated testing of black-box autonomous systems that have been applied across many domains.

I am currently on the academic job market looking for positions at the intersection of AI and sustainability.

Outreach on the Risks and Opportunities of AI

My research in safety-critical autonomy gives me a unique perspective into the risks and common pitfalls of deploying AI systems in high-stakes domains. I try to communicate these risks to non-technical audiences, seizing every opportunity to foster a broader understanding and awareness amongst key stakeholders. I enjoy discussing the challenges, best practices, and known limitations of AI systems with decision-makers in government (e.g. through HAI’s congressional bootcamp on AI, or as a member of the Wilson Center’s Pathway to AI Policy program), national security (e.g. at events hosted by CISAC or IST), and industry (e.g. my summer course on reliable AI through SCPD).

Hobbies and Interests

Beyond research I enjoy reading, spending time with friends and family, and being outside, especially rock climbing. It’s my dream to one day make a clean ascent of Freerider on El Capitan.