Peter David Fagan

I’m building Corca Health, an AI-assisted ADHD screening platform designed to support GPs with early detection and triage.

My broader work focuses on preventative medicine, secure health data systems, and safe, reliable AI for clinical use. I have a background in mathematics and computer science, with a focus on artificial intelligence.

Outside of work, I enjoy CrossFit, sea swimming, and hiking.

First-Author Publications

Peter David Fagan

Preprint (v2), 2026

Project Page

While often treated as abstract algorithmic properties, intelligence and computation are ultimately physical processes constrained by conservation laws. We introduce the Conservation-Congruent Encoding (CCE) framework as a unified, substrate-neutral physical framework for studying intelligence. We propose that information processing emerges when open systems undergo irreversible transitions, carving out macroscopic states from underlying reversible micro-dynamics. Generalizing Landauer's principle to arbitrary conserved quantities via metriplectic flows, we derive a universal bound for macroscopic computation. This yields physical metrics for intelligence and an operational analogue for consciousness, quantifying an agent's ability to extract work from the environment while minimizing its own dissipative dynamics. Applying CCE to the limits of physical observation, we model measurement as an active coarse-graining process rather than a passive projection. At the quantum scale, CCE recovers the Lindblad Master Equation, consistent with modelling decoherence as the dissipative exhaust required to record a measurement. Scaling to cosmological limits, we explore the hypothesis that gravity emerges as the macroscopic geometric footprint of these bounds. We show that, under this hypothesis, measurement-induced dissipation is consistent with a volumetric phase-space collapse, offering a dynamical route to the Bekenstein-Hawking area law. Equating the Landauer exhaust of this coarse-graining to horizon deformation outlines a limiting-case recovery of the Einstein Field Equations. Ultimately, by establishing a substrate-neutral link between thermodynamic dissipation, quantum measurement, and spacetime geometry, CCE provides physical constraints for understanding both natural and artificial intelligence.

This is a working manuscript that proposes a unified physical framework for studying intelligence. Several of the broader implications—particularly those bridging macroscopic information bounds with quantum and cosmological limits—are presented as formal hypotheses to be refined through extended proofs and empirical validation. The overarching goal is to anchor abstract computation in fundamental physical laws and, in doing so, establish rigorous, geometric constraints for the safe development of artificial intelligence.

Peter David Fagan

Preprint, 2025

Project Page / arXiv

We introduce a framework for applying keyed chaotic dynamical systems to encrypt and decrypt tensors in machine learning pipelines. This lightweight, deterministic approach enables authenticated inference without modifying model architectures or requiring retraining. Designed for privacy-first AI, this method provides a new building block at the intersection of cryptography, dynamical systems, and neural computation.

Peter David Fagan, Subramanian Ramamoorthy

Preprint, 2024

Project Page / arXiv

We introduce a new recurrent neural network layer that incorporates fixed nonlinear dynamics models where the dynamics satisfy the Echo State Property. We show that this neural network layer is well suited to the task of overcoming compounding errors under the learning from demonstration paradigm. Through evaluating neural network architectures with/without our layer on the task of reproducing human handwriting traces we show that the introduced neural network layer improves task precision and robustness to perturbations all while maintaining a low computational overhead.

Collaborative Publications

Rimvydas Rubavicius, Peter David Fagan, Alex Lascarides, Subramanian Ramamoorthy

Preprint, 2025

Project Page / arXiv

In this work, we introduce an interactive task learning framework to cope with unforeseen possibilities by exploiting the formal semantic analysis of embodied conversation.

The DROID Dataset Team

Robotics: Science and Systems (R:SS), 2024

Project Page / arXiv

In this work, we introduce DROID (Distributed Robot Interaction Dataset), a diverse robot manipulation dataset comprising 76k demonstration trajectories or 350 hours of interaction data, collected across 564 scenes and 86 tasks by 50 data collectors in North America, Asia, and Europe over the course of 12 months.

Open X-Embodiment Team

IEEE International Conference on Robotics and Automation (ICRA), May 2024

Project Page / arXiv

In this work, we introduce Open X-Embodiment, a comprehensive collection of robotic learning datasets and RT-X models. These datasets and models facilitate research in embodied AI by providing large-scale, diverse, and realistic environments for training robotic systems. The datasets cover a wide range of tasks and scenarios, enabling robots to learn complex behaviors through interaction with their environment.

Software

Peter David Fagan

Google Summer of Code, 2022

Code

This is the official Python binding for the MoveIt 2 library.