Born in Humble.Raised in Borneo.Schooled in Qatar.Building in Austin.

We are building a drone for Gulf conditions: sustained 50 °C ambient heat, fine dust, and contested radio where the control link is jammed. Existing platforms are adapted to the region; this one is designed for it. It is for clients in the Middle East who need routine inspection of pipelines, canals, power lines and solar plants to continue through summer, on schedule.

Three things break an ordinary drone here. Heat: we wrap the battery and flight computer in a phase-change jacket, duct air over it, and build the shell to carry heat out rather than hold it. Dust: motors are sealed and the electronics are potted. Jamming: the aircraft carries its whole mission on board, keeps flying it when the link drops, and finds its own way back to the dock.

I lead a partnership between the Sustainable Building Initiative at UT Austin and the Energy Club at the University of Ibadan, modeling how Nigeria reaches universal electricity access across all 36 states. The core is a monthly system-dynamics model of the electricity sector, tested against a 30-month backcast of primary NERC data across 33 falsification gates. It passes 22. We publish which 11 it fails and which conclusions survive: long-horizon supply levels are unreliable because the maintenance-backlog stock saturates, but the ordering of investment pathways and the collapse, access and equity results hold.

The spatial layer runs across 36 states and 6 geopolitical zones, tracking an equity Gini alongside cost. It shows that least-cost and equity-optimal capital allocation converge: expanding access to the underserved north carries no efficiency penalty. That result is one of the ones that survives the gates.

The work is ongoing and this page shows only what the team has agreed to release ahead of Geneva. Findings are being prepared for presentation at the 15th UN Forum on Business and Human Rights, Geneva, November 2026.

Pavement distress surveys at the lab were run by strapping a GoPro to the tailgate of a truck and driving the road. The footage was usable but the process wasn’t. Frames blur at speed, the camera’s optics cap what you can resolve, and every hour of video needs a person to scroll through and annotate it. There were never enough people, so the backlog grew faster than the data got processed.

I proposed replacing it: a UAV imaging pipeline for the aerial capture, and a trained visual learning model to do the annotation. To show what that would look like before it existed, I built a full mission simulation, a test road with generated distresses, an aircraft that flies the lines and triggers the camera on distance, a simulated onboard detector whose confidence tracks crack width and ground sample distance, and an orthomosaic stitched from the captured frames. I took it to Dr. Zornberg and he funded it: a $13K aircraft package, and permission to staff the team from my own organization. It’s now 14 people comparing pavement distress on roads built with wicking-and-reinforcing geotextiles against untreated control sections.

Training the model needed compute the lab didn’t have, so that became the next problem I was handed. I was given a $15K budget, specified every component myself against what ML training actually demands in RAM, CPU and GPU, and built the machines for the lab.

The lab’s wicking-geotextile experiments were logged by hand: walk in, read the scale, write down how much water had moved through the fabric into the beaker. That put a person in the loop for weeks at a time. Readings had to happen at the same hours every day to stay comparable, which meant they happened only when someone could reliably be there, and almost never overnight. The experiment ran continuously. The observations didn’t.

I strapped a GoPro to the glass humidifier wall and wrote a Python script that extracts a frame every minute and reads the weights off the images. No lab visit required.

The rig is a closed mass balance inside a temperature- and humidity-controlled chamber, so evaporation can’t hide a loss: a full beaker zeroed on a scale feeds the geotextile, and the water that wicks across collects in a catch beaker weighed at the end. It has to add up to the mass the scale watched leave.

The lab now records 288× more entries, and it records them through the night. The method was adopted lab-wide for all similar experiments, turning point-sampled graphs into continuous ones.

Accredited by the UNESCO Chair on Environmental Law and Sustainable Development at Hamad Bin Khalifa University to attend the 15th UN Forum on Business and Human Rights at the Palais des Nations, November 23–25, 2026, on behalf of the Chair. Carrying research on equitable electricity investment in Nigeria, developed in partnership with the University of Ibadan, into the Forum’s 2026 agenda, which marks fifteen years of the UN Guiding Principles on Business and Human Rights.

It won’t be my first time in rooms like that. At the ASCE Legislative Drive-In I was the youngest participant, briefing legislative staff on three infrastructure bills, with four U.S. senators on the schedule.

At the TCEQ I ran water quality tests on state treatment systems, checking compliance against Texas safety regulations, and sat in on the agency’s regulatory review of air quality, landfill, and wastewater programs.

The part that stuck was watching the legal and engineering teams work against each other and together. Regulation isn’t written by engineers alone, and the public doesn’t hear from lawyers alone. Seeing where those two meet, and how government contracting actually runs underneath it, shaped how I think about the client side of this work.