I build, scale and recover the engineering organizations behind software that has to work.

As Head of Software Engineering for cockpit electronics, a field issue surfaced on the Ford Bronco S-family: intermittent display de-illumination on units already in production. Safety-critical, escalated by a demanding OEM, with recall risk in the millions.

The task: root-cause it and ship a fix without a hardware recall. The fault did not reproduce easily on the bench, and it lived in a concurrency marginality of the dual-processor design: non-atomic read-modify-write operations on shadow RAM in the init routine, racing the CAN interrupt handler, behind an integrity check too weak to catch the silent corruption.

I took over the task force lead in front of the OEM and stood up a global SWAT team within 48 hours across Mexico, Germany, Singapore and the United States. I ruled out a hardware respin; the fix was pure software: the arbitration flag moved into its own volatile-qualified, naturally aligned memory cell, and every read-modify-write was wrapped in a critical section for atomicity against the CAN ISR. We validated with HIL and in-vehicle testing across sites, and I kept the team motivated to sign-off under sustained OEM pressure.

The GM T1xx global cluster program carried heavy technical debt against a fixed OEM date, with 50+ engineers across the US, Mexico and Bulgaria plus Luxoft contractors in Ukraine and Guadalajara.

The platform was inherited, with fundamental flaws not priced into the award. Safety logic and rich graphics also could not share one processor without compromising one, and the date would not move.

With the architects I reviewed the inherited platform and raised its flaws early so they got resourced, not absorbed. I ran the master sprint plan that protected velocity and quality toward feature-complete, coordinated the HMI and graphics team on the NXP MPU to it, and managed the team that kept graphics and the AUTOSAR Classic safety side on the Cypress MCU in lockstep over the protected UART bridge.

A first-of-its-kind aftermarket telematics product for legacy vehicles, built from nothing: no existing functional modules, an aggressive deadline, and a five-engineer hardware and software team.

Everything was new at once: an OBD-II hardware module, multi-vehicle decoding, GPS, cellular, over-the-air programming and RF signaling had to cooperate on a constrained target, against a schedule that assumed reuse that did not exist.

I led the product concept and technical strategy for the new line, led the team that architected the platform on a Micrium uC/OS RTOS with third-party libraries for multi-vehicle decoding, led the location tracking and remote door-lock logic in embedded C and C++, and coordinated a dedicated validation group proving GPS accuracy and RF integrity under field conditions, not lab conditions.

Chrysler PODS units were going brain-dead during vehicle-level reflashing, halting production.

The corruption happened mid-reflash, deep in memory, and left no obvious trace at the application level.

Traced it through deep memory analysis to the first bootloader section being partially erased during the J1850-over-CAN sequence, and shipped a memory boundary protection fix.