Johns Hopkins University Applied Physics Laboratory
He is a principal research engineer, project manager, and section supervisor at APL, and a faculty member at Johns Hopkins, applying systems engineering methods to complex challenges spanning defense, space, biomedical, and learning systems — building scalable data systems, human–machine teaming solutions, and the programs that grow the people who do this work.
The through-line is systems engineering plus measurement, applied wherever capability must be built — in a reconstruction pipeline, a clinical dataset, a research team, or a learner. The same discipline that maps a cubic millimeter of cortex also builds the archive that shares it, the metrics that judge it, and the programs that staff it.
Mapping brains at synaptic scale — automated reconstruction pipelines, petabyte-scale community archives, and the query tools that turn image volumes into testable circuits.
Computer vision, graph analysis, and cloud infrastructure that scale from laptop prototypes to petabyte pipelines — including algorithms that borrow their architecture from biology.
Two decades of systems engineering for national-security sponsors — from satellite systems as a responsible engineer to decision-support prototypes for hard operational problems.
Treating human capability as a system parameter — something you can specify, measure, and improve with the same discipline applied to any engineered system.
Cohort-based pathways into research careers — approximately 500 students mentored through programs built to widen who gets to do science.
The same measurement discipline turned on clinical data — analytics leadership for multiple sclerosis and COVID-19 research at Johns Hopkins.
For students and mentees — the programs, and how to get in touch →
CIRCUIT — the Cohort-based Integrated Research Community for Undergraduate Innovation and Trailblazing — is a nationally recognized research and mentoring program he founded and led at APL from 2017 to 2023. It served 220+ students from high school through graduate school, across technical areas from underwater to space systems, and its cohorts contributed to work published in Nature. Co-author of the peer-reviewed CIRCUIT model paper (ASEE 2023).
The College Prep Program at APL, which he co-founded in 2009 and still co-directs, is an all-volunteer summer program supporting underserved students through academics, applications, essays, and financial aid, with longitudinal support afterward. A further 220+ students served, with more than 90% on track for four-year degrees, on 500+ volunteer hours a year.
On the learning-engineering side, he leads the development of the LENS concentration — Learning Engineering for Next-Generation Systems — within the Johns Hopkins M.Ed. in Learning Design and Technology, the program directed by James Diamond at the School of Education, where Gray-Roncal is a Visiting Assistant Professor. The work applies systems engineering principles to learning in operational environments including healthcare and defense.
He teaches Frontiers in Neuroengineering (JHU Engineering for Professionals), Agentic AI Systems (JHU Lifelong Learning), and Mapping the Brain: An Introduction to Connectomics (Johns Hopkins University). The connectomics intersession course was consistently among the highest-rated in the Whiting School.
Functional connectomics spanning multiple areas of mouse visual cortex ★
Using machine learning on clinical data to identify unexpected patterns in groups of COVID-19 patients ★
An Immersive Curriculum to Develop Computational Science and Research Skills in a Cohort-Based Internship Program ★
A framework for rigorous evaluation of human performance in human and machine learning comparison studies ★
The Brain Observatory Storage Service and Database (BossDB): A Cloud-Native Approach for Petascale Neuroscience Discovery ★
DotMotif: an open-source tool for connectome subgraph isomorphism search and graph queries ★
Benchmarking Human Performance for Visual Search of Aerial Images ★
A community-developed open-source computational ecosystem for big neuro data ★
Quantifying mesoscale neuroanatomy using X-ray microtomography ★
Saturated Reconstruction of a Volume of Neocortex ★
Graph classification using signal-subgraphs: applications in statistical connectomics. ★
Full publication list → · 3,009 citations · h-index 21 · i10-index 32 (Google Scholar, August 2026)
Programs led or co-led at APL, sponsored by NIH, DARPA, IARPA, and DoD.
Scaling circuit analysis and reconstruction
Petabyte-scale data storage and community collaboration for connectomics and precision medicine
Community standards for high-resolution connectomics
Low-shot learning algorithms inspired by insect neural connectivity
Algorithms for cubic millimeter mammalian neuroscience; scene parsing methods for operator-focused environments
Reproducible cloud processing pipelines for neuroscience
Translating biological neural computation to next-generation ai algorithms
Quantitative measures of operator capability in complex operational contexts
One claim runs across the domains: what people can do inside a system is a designable, measurable property of that system — and treating it that way changes what you build, what you measure, and who gets to do the work.
Systems don't fail because the technology breaks. They fail because someone couldn't do what the system required.
Every complex system depends on what people can do inside it. That dependency is measurable, designable, and too important to leave to chance. Capability lives at the interface between what a system requires of its operators and the impact the system has to deliver — and when that interface is wrong, the gap may originate in human development, in system design, in organizational performance, or in the interaction among them. Telling those apart is itself a measurement and engineering problem.
Casebook, First Edition, August 2026
A curve can turn a 40 into a B. A transcript can certify a degree. A benchmark can certify a model. None can tell you whether the learner, the clinician, or the operator can actually do the thing when the stakes are real — only a demonstration can.
The argument that professional competence has to be measured by functional behavior — a work sample scored against a real task — not inferred from a grade, a credit hour, or a test score. The same standard doubles as a fairer gate: let the demonstration, not the pedigree, decide.
Draft, August 2026, with James Diamond
"More training" is the answer we reach for when we have not asked what capability is actually missing, or where in the system it lives.
You cannot lecture a pilot out of a single-sensor design, audit a hospital into safety with a metric staff have learned to game, or brief a crew past a manning shortfall. When the diagnosis is wrong, the intervention is theater.
Measurement discipline belongs on the people running an experiment as much as on the instruments they run it with.
BRAIN CONNECTS, in press
"Not like this, not yet" is a result a capability engineer must be able to deliver.
Reading a constraint space accurately enough to recognize that external, non-technical factors have narrowed it past the point where a capability goal is realistic is itself a valid project outcome — documented, defended, and used to guide the work. It is not a failure.
Capability without agency is automation.
The goal is not operators who comply with a system specification but operators who can perform, adapt, and lead when the system encounters conditions it was not designed for.
Infrastructure, tools, and teaching artifacts — most open source, all built to be used by someone other than their author.
The Brain Observatory Storage Service and Database — petabyte-scale storage and delivery for volumetric neuroscience, serving the connectomics community as shared infrastructure.
A query language and engine for finding subgraph motifs in connectomes — ask structural questions of a brain network without writing a traversal.
Training and mentorship for nanoscale connectomics: ten technical units from electron-microscopy basics through NeuroAI analysis, a curated journal club, and hands-on work against real datasets including MICrONS and FlyWire. Built for three audiences at once — students, the mentors supporting them, and programs adopting the model.
An AI mentor working from an operational knowledge corpus — symptom to fault to corrective action, plus the tacit cues experts never write down — that probes a learner through a live scenario at a chosen level of expertise. It assesses on unfamiliar scenarios, because repeating a known pattern is not transfer.
Grades an argumentative essay twice — once from the finished text, once from the transcript of the AI dialogue behind it — and reports the gap between them. Every score is anchored to a verbatim quote, and criteria a model should not settle alone route to the instructor, whose override becomes calibration data.
Next-token prediction explained at five depths, so a reader can take the level they can actually use rather than the one the author found interesting. The build is written up beside it as a worked example of designing for a capability instead of for coverage.
A soap-bubble physics sandbox: shape a wand, watch the film billow and deform in moving air, then run the whole thing backwards — fit an ellipse to a photo of a real bubble and read the wind speed back out of the shape it left.
A shared reading list, tool directory, practice library, events calendar, and community roster for people doing learning engineering.
Johns Hopkins APL News
Announcing BENCHMARK, a $1.3M NIH grant to build community standards for large-scale structural brain data, integrated into bossDB.
We will significantly amplify neuroscience research investments by ensuring data and analysis products are comparable and reproducible.
Johns Hopkins APL News
Named co-lead of APL's contribution to two NIH BRAIN CONNECTS projects, bringing scalable cloud pipelines, systems engineering, and machine learning to high-throughput connectomics.
JHU Engineering magazine
A feature on BENCHMARK and bossDB, and on making petabyte-scale brain data shareable across the BRAIN Initiative community.
Johns Hopkins APL News
bossDB joins the AWS Open Data Sponsorship Program, opening petascale neuroimaging data to the public.
This relationship will catalyze the democratization of data access and accelerate scientific exploration by researchers and members of the public.
Johns Hopkins APL News
Led the data science team that triaged records from 1,182 COVID-19 patients and built an algorithm predicting patient trajectory two weeks after admission.
Johns Hopkins APL News
Launch of the Johns Hopkins Precision Medicine Analytics Platform (PMAP).
PMAP gives researchers access to data, a virtual collaborative workspace and tools to help develop new algorithms that can ultimately improve clinical decisions and patient care.
Johns Hopkins APL News
The bossDB launch announcement.
As neuroscience datasets continue to grow in size, this is an important example of collaborative experimentation at the forefront of neuroscience, combining diverse expertise drawn from many research groups.
Johns Hopkins APL News
Awarded jointly to Will and Karla Gray-Roncal for co-leading APL's all-volunteer College Prep Program.