Field
Every observer has limited bandwidth and a world larger than it. The research asks one question across many settings: given partial information and finite resources, where should a system look next — and how does what it can perceive change as it learns?
That question has been asked of underwater vehicles following algal blooms, drones counting fruit and detecting disease in orchards, robots mapping coral reefs, fault scarps, and precariously balanced rocks, digital twins for lunar mission planning, and — most recently — the kernel itself: the mathematical object that determines which distinctions an agent can represent at all.
A web-based digital-twin ecosystem for Earth and space science — annotation, semantic mapping, decision support. Its dual use as a 1:1 template engine for artisans in Odisha is documented in Engineering Mythology.
An open, cloud-enabled testbed for UAV education and research, built for the NSF Student Cyber-Physical Systems Challenge — simulation, autonomy, and a collaborative design studio for field robotics.
An autonomous underwater vehicle for persistent coral-reef monitoring and bathymetric mapping, operating in tandem with the R/V Karin surface vessel — a heterogeneous team with coordinated dock and undock.
Shakebot, a low-cost open-source robotic shake table, and the Virtual Shake Robot — simulating how precariously balanced rocks overturn, as constraints on past ground motion.
High-resolution terrain databases for mission simulation, planetary-scale feature detection and mapping, event-based vision for autonomous robot operations, and lunar lander descent imaging.
A variational framework in which the kernel — what an agent can distinguish — is itself the dynamical variable, governed by Maximum Caliber. Four papers in 2026, the latest applied to planetary surface graphs.
Marine robotics (USC, MBARI, 2006–2014). Data-driven sampling of dynamic ocean features with AUVs, gliders, drifters, and surface vehicles; bloom-trajectory prediction for mission planning; mixed-initiative multi-robot field experiments in Monterey Bay and the Southern California Bight.
Agricultural robotics (Penn, 2014–2018). Automated monitoring for precision agriculture; deep-learning fruit counting across orchards; close-range remote sensing for citrus greening detection; aerial phyto-biopsy; two patents.
Before robotics (2004–2006). Large-scale distributed enterprise software at ThoughtWorks, built with agile methods — which is where the habit of shipping iteratively and documenting everything comes from.