01
Neuromorphic & cognitive computing
I research computation through dynamics to build neurally inspired artificial intelligence.
Computational neuroscience, dynamical systems, electrical engineering and machine learning — used together to find mechanisms of coordinated computation, and to put them in hardware.
Rebound element → column → spatial code
I am a senior member of technical staff in Neuromorphic Computing at Sandia National Laboratories, where I lead projects on physics-based computation in emerging hardware, graph neural networks for scientific computing, and physics-informed neural networks. As principal investigator I direct a portfolio of roughly $4M across four grants, and I lead cross-functional teams of four to nine researchers.
My work is grounded in the study and functional abstraction of biological circuits. During my PhD in Computational Neuroscience at Boston University with Michael Hasselmo, I built models of neocortical microcircuits that learn from temporal error — self-supervised prediction rather than reconstruction. Before that I was a neural modeler at eCortex and a research software engineer at the Conte Center for Systems Neuroscience.
Recent
2026-07
Co-instructing Neuromorphic Computing at the University of New Mexico this autumn.
2026-03
Became a senior member of technical staff in Neuromorphic Computing at Sandia, after three years as a postdoctoral appointee.
2026-03
Graph Reservoir Networks for Prediction of Spatiotemporal Systems presented at NICE 2026; patent application 19/550,784 filed in February.
2026-02
Invited talk at Georgia Tech — “Intracellular dynamics for mixed signal neuromorphic computation.”
2026-01
Three invited talks in January: the Energy Consequences of Information workshop in Santa Fe, Flagship Pioneering in Boston, and GreatSky.ai in Boulder.
2025-07
Received Sandia’s Distinguished Mentorship Award for work with graduate students.
Three threads
What I am working on
These are three stages of one program: neural recordings reveal the computations biological systems must perform; circuit models identify the dynamics that make them possible; physical neural networks test whether those mechanisms can become efficient, adaptive computing systems.
I also apply this dynamics-first approach to scientific machine learning, including graph reservoir networks and physics-informed models of spatiotemporal systems.
Thread 01
Physical neural networks
Let the device dynamics do the computing instead of simulating them — ferroelectric capacitors, analog in-memory arrays, temporal elements.
6 papers →
Thread 02
Predictive microcircuits
Hierarchical, feedback-driven models of neocortex that learn from temporal error — plus the rebound dynamics and reservoir surrogates that come out of them.
7 papers →
Thread 03
Spatial coding & navigation
Egocentric boundary and speed codes across retrosplenial, parietal and entorhinal cortex — and the GPS-free mapping and localization they support.
8 papers →
Selected work
All 23 papers →02
Predictive Learning by a Burst-Dependent Learning Rule
03
Egocentric Boundary Vector Tuning of the Retrosplenial Cortex
Working on dynamics, devices, or both?
I welcome conversations about neural dynamics, neuromorphic hardware, predictive learning, spatial cognition, and mentorship.