Coming soon once public release is approved.

The past couple of months have been challenging on both the technical and mental front. I returned to working on PMHS-related projects, this time in closer proximity to the people who donated their bodies to science. The meticulous preparation and the stress embedded in test setups are all in service of data, and with PMHS, some of that data takes considerable time to materialize. In an effort to extract preliminary insight, I built a small piece of software to isolate the skeleton from DICOM images, which introduced an entire learning curve at once: a new data format, new knowledge in human tissue, organs, Hounsfield units and scan rendering, new Python libraries, and a new quality checklist to validate the work.
When the task was first pitched to me, curiosity and excitement got the better of my judgment. Once I started, however, I quickly realized I had no training, experience, or prior exposure that would prepare me to examine a deceased human from multiple angles, let alone to set aside their humanity and treat them as a data repository.
I had a few conversations with peers and collaborators on the subject. Some employed dissociation, others claimed genuine indifference, and a few were candid about the weight of it. The consensus seemed to be that I needed to arrive at a place where the scans were pure data and the focus was purely technical.
Having spent the last four years working with datasets across different contexts, including machine learning, the ethical dimension was never far from my thinking: the sourcing and use of databases, the responsibility of the R&D engineer to remain transparent throughout the process, and the reality that capability does not confer justification, particularly in an environment where foreseeable mistakes persist precisely because the threshold for concern is treated as negotiable.
One argument that circulated was that when a person consents to donating their body to science, they relinquish control over its specific use. Their family does not necessarily know where the body will end up, and anonymity is maintained through a de-identification pipeline. Under that logic, whether the body is used for a training dataset, a crash test, or a classroom exercise is immaterial. The consent is total and the purpose is scientific.
That logic, however sound, did not resolve my agitation. I have to credit the project lead, R., who listened to my concerns without minimizing them, shared what had worked for him, and gave me adequate time and room to find my own footing.
In the end, I got through it, delivered feedback-incorporated updates, and produced a robust final version. Several things made that possible:
- Rather than starting with the lab's own data, I used the Visible Human Project from Harvard Dataverse to build and troubleshoot the program. The repository provides anatomical components rather than complete bodies, which gave me gradual exposure and allowed me to develop a proof-of-concept before working with a full human scan.
- I abandoned my peers' coping strategies, because dissociation and soldiering through did not hold up for me. I accepted that after roughly two hours of sustained work, I needed to stop, and sometimes to cry, and I gave myself that without reticence. I paced myself, took breaks. I already had a meditation practice and that came in particularly handy as well.
- I gave the PMHS subjects names drawn from thinkers and scientists I admired growing up. That helped me stay connected to their agency in choosing to donate, the generosity that made the work possible, and the purpose behind it all.
- I kept other projects running in parallel to shift the cognitive register and interrupt the accumulation of that specific kind of mental weight.
- After hours, I brought Sonny to my desk, my energetic and mischievous beagle, who imposed breaks, demanded walks and sprints and attention, and provided what was arguably the most effective emotional reset of the entire process, carrying me through to the stage where I was working on full-body scans and examining the intricate details of the skeleton before and after impact.
I cannot say with certainty that I would approach a similar assignment with the same enthusiasm, but I know I would get through it. I also came away with a more honest sense of how far naive excitement and scholarly motivation toward a totally novel field can carry me before the full reality of a project catches up. What I am most grateful for is that I surfaced the difficulty early, had professional support alongside me, and found a way through that was genuinely mine.

The test was part of an ongoing research program on the Integrated Crashworthiness Safety approach of future Electric Vertical Takeoff and Landing (eVTOL) aircraft. NIAR-AVET evaluated the crashworthiness performance of eVTOL battery packs and their surrounding structure during a free fall of 50 feet. The primary objective of this test program and simulation study was to identify the behavior (Structural, Thermal, and Electrical) of the battery pack during emergency landing conditions in order to provide the FAA information that may be used to define future requirements and how its performance will impact the selection of composite materials for the construction of an airframe capable of providing an adequate level of safety.

"By the way, do you know where we can potentially recycle a 750 lbs battery?"
I did not, off the top of my head. Being intelligent across different fronts does not, as far as I am aware, come with supplier network psychic powers, a conviction my chief scientist and main collaborator seemed to hold with inexplicable confidence.
What made the timing of that question particularly memorable was what had just preceded it. In the same conversation, my chief scientist had casually shattered my worldview by introducing me to ChatGPT and demonstrating its capabilities live, including making it generate in under two minutes the Python numerical homogenization code I had spent eight months of my Master's degree building. That being said, one worldview collapses so another can rise, and I had already been preparing for the AI wave by taking a neural network course taught by my new advisor, and investigating machine learning applications within my PhD research, so the blow landed on reasonably prepared ground.
Within a month, I was onboarded onto an additional project supporting an eVTOL battery 50-foot drop test, handling testing and data acquisition alongside D. and M., with a team of undergraduate researchers and engineers. It was our first battery drop of any kind, and it remains to this day the largest and heaviest battery I have encountered in any professional capacity (2025 update: still holds).
The safety footprint was appropriately expanded for the occasion. A fire marshal, a large water tank, and a standing fire crew were all on site. The test took place in an open field in ice-cold weather, and the lighting was, once again, a genuine problem for DIC purposes. No explosion however. The test yielded better results than anticipated, and the learning curve was steep enough to make it clear we would be taking on more battery-related work in the future.
As for the recycling question, after several phone debates, a handful of concerned organizations calling back to ask what exactly we were doing with a battery that size and whether anything would seep into the ground (answer: proprietary, and no, we have a full concrete basin for testing), I eventually found a contractor willing and equipped to take on the remnants. Shoutout to Chad H., who did not flinch. (2025 update: I still get asked where to recycle large batteries, but now my answer is ready and that excellent contractor still is my go-to contact).
A side note that I feel compelled to include: if you ever find someone gracious enough to handle a large battery load, do not mix in metal or composite scraps, and do not be difficult with the people who make pre- and post-test cleanup possible. The goodwill of suppliers who go above and beyond is finite, and the kind of work we do makes us acutely dependent on it. What makes the occasional lapse in basic professional courtesy particularly bewildering is that the nature of our own work puts us on the supplier end of that same dynamic regularly enough that we should know better.
Some collaborators half-joke that a certain flavor of dismissiveness correlates with academic credentials and with the kind of identity investment that comes from living too exclusively inside the academic bubble, which is its own irony, given that the work itself demands positioning across disciplines, industries, and professional cultures that have no patience for puritanical boundaries.
I am planning to get my PhD the following year and will report back on the dickery coefficient. (2025 update: I got my PhD in late 2024 and I have since unlocked a kaleidoscopic and simultaneous distaste for pompous academic behavior, performative intellectualism, condescension toward academic AND non-academic knowledge, and the reflexive glorification of blue-collar work. Congratulate me, I am not automatically in anyone's good graces, which I consider a predictable outcome.)

As part of an ongoing fifteen-year research program on the Crashworthiness Certification of Composite and Metallic Aircraft Structures, a drop test was conducted at the new AVET facilities with the support of the Injury Biomechanics Research Center of Ohio State University. The OSU team provided their support and expertise with the three PMHS passengers.
The test carried four concurrent objectives: to verify and validate the finite element modeling techniques underpinning virtual crashworthiness certification of future aircraft programs; to collect injury data from the seated PMHS occupants and compare their biomechanical response against FAA Hybrid III 50th percentile ATDs; to generate datasets for the verification and validation of finite element human body models intended for injury prediction; and to characterize the injury potential for an occupant in a supine position on a medical stretcher, directly supporting the human body modeling activities of the Office of Naval Research I-Predict program.

I don't remember the last time I was so excited to come to work that I would be there very early in the morning (except during a LEAP engine certification program at Safran, which was my first time participating in a certification process). I was consistently arriving at work before 5:30 a.m. and still finding colleagues already on site. Between DIC setup under challenging lighting conditions and ATD preparation, I managed to shadow the OSU biomechanics team and gain firsthand exposure to their sensor placement philosophy, their handling protocols for the PMHS occupants, and their calibration and data acquisition procedures.
The person who shaped my understanding of the instrumentation side most directly was Jonathan Conklin, who headed AVET's calibration lab at the time and held full responsibility for the ATDs from pre-test calibration through post-test teardown and maintenance. Data acquisition was, as it should be, treated as a critical and fragile component of the entire effort, and the collective frustration on site when R&D personnel unfamiliar with its requirements would walk over wiring or mishandle DAQ equipment was considerable and entirely justified.
Working in close proximity to cadavers was at turns awe-inducing and disorienting, in the way that any sustained encounter with the fundamental materiality of the human body tends to be. What proved equally striking, and what has stayed with me longer, was the parallel immersion in the state of ATD technology and the extent of its limitations. The regulatory standard for aviation emergency landing testing remains a male dummy whose design is rooted in decades-old anthropometric data. The existing female ATD is not derived from female anatomical data at all, but is a geometrically scaled-down version of the male model, carrying none of the physiologically distinct properties of female musculature, spinal geometry, bone density, or injury response. A true 50th percentile female ATD does not exist in regulatory use, and the injury data bears that out in ways that are neither subtle nor acceptable.
Through daily calibration routines and sustained conversations with Jonathan, a picture of the gap between what the PMHS data was capturing and what the ATDs could actually represent began to take shape. His methodical precision and deep understanding of every sensor, every tolerance, and every procedural boundary made those exchanges genuinely instructive, and the questions they raised about how PMHS-derived datasets could drive improvements in both ATD modeling and fabrication remain as consequential as the test data itself.
We joked, but not really, that even the regulatory standard for the male ATD does not reflect the average data of the male population as of now, age or health wise. And while testing remains expensive, and PMHS procuration does not guarantee finding average measurements reflecting of the intricacies of the humans of today, this is yet another arguent for certification-by-analysis, and the importance of augmenting simulation capabilities and incorporating them to the certification lifecycle: Libraries of accurate human body models and investigations of crash related simulations involving them would bridge some of the gap that sole physical testing doesn't get right, and allow to dimension better for a more inclusive scope of humans. That's why projects such as the I Predict (SwRI has an excellent introduction to the full concept) are crucial step towards the right direction, and deserve keen attention and serious funding, that's why body donation is priceless in the way it would inform the next generation of inclusive regulations.

The FAA Hybrid III 50th percentile male ATD, which remains the regulatory standard for aviation emergency landing testing, is a derivative of a design developed based on U.S. anthropometric data from 1976, codified in federal regulation with drawing packages dated as late as 1998. The existing female ATD, a 5th percentile figure introduced in 1966, is not derived from female anatomical data but is instead a scaled-down version of the male model, carrying none of the physiologically distinct properties of female musculature, spinal alignment, bone density, or injury response. A true 50th percentile female ATD does not exist. The consequences are measurable: a 2019 University of Virginia study found that women are 73% more likely to be injured in a frontal crash than men, controlling for age, height, BMI, crash severity, and vehicle model year, and NHTSA data indicates women sustain roughly 80% more injuries to the neck, chest, and pelvis. The THOR-50M, NHTSA's next-generation male ATD with improved biofidelity and expanded instrumentation, has been under development since 2009 and as of 2024 remains in regulatory validation. An equivalent female THOR is not yet on a confirmed regulatory timeline.

On April 10, 2010, a Polish Air Force Tupolev Tu-154M (registration 101) crashed near Smolensk North Airport in Russia, killing all 96 people on board. The official investigation, summarized in the MAK report, attributed the accident to pilot error under conditions of severely reduced visibility during approach.
NIAR was subsequently contracted to conduct an independent numerical analysis of the accident, with two specific objectives: first, to evaluate the structural damage predicted by a finite element model of the Tu-154M against the damage documented in the MAK report; and second, to assess the survivability conditions of each occupant by comparing the injury outcomes derived from the simulation against the known injuries reported in the investigation record.

The crash involved a collision with a birch tree, which meant that for a structural reconstruction to accurately proceed, a thorough wood characterization campaign had to be conducted, covering both physical testing and simulation. Wood is not exactly what one expects to be researching in an aircraft institute, but by that point I had learned that NIAR's project portfolio operates on the principle that if the investigation demands it and the facilities can support it, the work gets done regardless of how far outside the domain it lands. Wood was unusual, but it was not the strangest material we ever characterized. That distinction belongs elsewhere, and involves skin and bone, which is a story for another entry.
What I had also learned by then was to embrace the learning curve as a structural feature of the work rather than a quirky punctual circumstance, and to treat a rigorous literature review as the actual foundation of every claim. The full team was involved in the FE reconstruction and verification, but the wood investigation itself was concentrated within my immediate group. I had the good fortune of working alongside D. on the birch tree identification, sourcing, mechanical testing, simulation, and the development and study of the MAT_143 material card. It consumed roughly a year of focused effort, run concurrently with my PhD qualifier preparation and coursework in theory of plasticity, and it remains among the work I am most proud of.
The project also surfaced a question that was raised but never formally resolved: whether the findings from these contained sub-investigations, particularly material characterizations conducted for a specific contract, should be published independently given their potential broader value. The literature review made the case starkly: the intersection of rigorous material characterization and wood is thin, and the available scholarship rarely delivers the level of detail required to construct a fully parameterized, generically applicable wood material card for simulation. The gap was real and consequential.
Given my lack of prior experience with wood and the curiosity that tends to follow me into every new domain, the project pulled me into the world of woodworking well beyond the lab. I spent time with local Wichita practitioners ranging from hobbyists and tradespeople to artists who treat wood as a primary medium, partly for the project and partly because I could not help myself. None of them had any awareness of what finite element analysis could offer in terms of predicting failure regions, guiding joinery decisions, or estimating the longevity of a given design in a given species, but several were immediately receptive to the idea when it came up, including to the notion of paying for material characterization of their wood of choice alongside a simulation-informed design review.
One person in particular made that entire side of the investigation possible in a way that goes beyond acknowledgment in a project report. Kody Ramsey opened his workshop without reservation, fielded an unreasonable volume of fired-up questions with patience and generosity, gave me something no textbook or material database could: a genuine feel for wood as a physical material with character, grain, resistance, and memory. That kind of grounded, tactile knowledge turned out to matter in ways that showed up directly in how we approached preparing the specimens, clamping conditions and simulation work. Most importantly, that unassuming, funny, kind, and dedicated man casually told me, as if merely seeking confirmation, what was arguably my greatest scientific epiphany of 2019: wood is among the most extraordinary and statistically improbable materials in the observable universe. Supernovas, comets, asteroid belts, and planets are abundant with minerals, heavy elements, and even exotic compounds, but wood requires biological life and a specific evolutionary emergence of plant tissue, a biochemical architecture that arose only once in Earth's own history. No planet yet discovered has shown any evidence of conditions capable of generating anything close to wood.
In addition to that, Kody was immediately receptive when the idea of simulation-informed design came up, to the point of genuinely considering paid material characterization of his preferred wood species alongside a design review, which made the market gap visible even from within a crashworthiness lab. It proved to me, once more, that intersectionality in science is a practical virtue, that overspecialization uncoupled from genuine curiosity about radically different ways of knowing and making is a form of intellectual impoverishment, rather than curation, and that science in the service of people is at its most meaningful.

The NASA Advanced Composites Consortium (ACC) High Energy Dynamic Impact (HEDI) program was a multi-institutional effort uniting NASA, the FAA, major aerospace OEMs (Boeing, Lockheed Martin, GE Aviation, UTC, Spirit AeroSystems), and academia -including Wichita State University and NIAR as named consortium members- with the goal of developing and validating Progressive Damage and Failure Analysis (PDFA) methods for high-velocity impact events relevant to aircraft certification, specifically fan blade containment, bird strike, and fuselage shielding scenarios.
The program's validation methodology followed a building block approach, which required coupon-level material characterization to populate the property inputs feeding PDFA models in LS-DYNA (MAT162, MAT261, MAT213, and Peridynamics formulations). This material characterization layer is where the present contribution was situated.

Upon joining the Materials Research Group at what was then the Computational Mechanics Lab (now AVET-NIAR), on of my initial assignments required me to integrate into an ongoing coupon-level experimental campaign operating under the NASA ACC HEDI program, where I contributed alongside an established team. The group was running mechanical testing on an MTS load frame instrumented with a Digital Image Correlation system, a full-field optical strain measurement approach particularly well-suited to capturing the heterogeneous deformation behavior of composite specimens that point-based extensometry would either miss or misrepresent. My contributions within this team effort spanned specimen documentation, raw data extraction and reduction from DIC outputs, translation of the measured strain fields into material property inputs for LS-DYNA material card development, participation in card verification and validation exercises, and co-authoring portions of the deliverable reports submitted to the consortium.
Beyond the technical learning curve, what stayed with me most was the leadership of Adrian "Lemonchete" Gomez Fernandez. Working under Dr. Raju's expertise on high-rate material characterization was eye-opening in its own right, and the fast ramp-up across tooling, DAQ systems, and software packages was its own education, but Adrian's ability to navigate the human side of technical work left a more lasting impression. He held a rare balance: solid engineering judgment, genuine approachability, the willingness to push back when needed, and a grounded instinct for keeping ambitious work tethered to feasible outcomes, all while managing a team drawn from genuinely different backgrounds and cultures. He was the reason I chose to continue with the materials research group when I started my PhD, and his departure left a noticeable gap that no one could quite fill.