About gradientdescent.dev
gradientdescent.dev is my public engineering notebook for learning by building.
I'm Weldon Malbrough, a longtime software builder and retired U.S. Army officer working from a home lab in south Louisiana. gradientdescent.dev documents my progression deeper into electronics engineering, embedded systems, edge computing, and high-performance computing - especially where hardware and software meet.
The work ranges from simple circuits and embedded firmware to motion-control systems, custom electronics, GPU inference, and intelligent machines operating at the edge.
Build. Measure. Understand. Iterate.
The name describes the method.
Build something. Instrument it. Measure what actually happens. Understand the gap between expectation and reality. Then build the next version better.
That approach applies whether the project is a circuit on a breadboard, Rust firmware running on a microcontroller, a precision motion platform, or an inference workload running on an NVIDIA GPU.
Failures, unexpected results, and wrong assumptions are part of the record. The goal is not to present every project as polished or novel. It is to document the progression from understanding components to designing increasingly capable systems.
Work labeled Research is held to a higher standard: explicit methodology, preserved evidence, provenance, and reproducibility.
Technologies
The technology changes with the problem, but current interests include:
Electronics & embedded systems
Analog and digital circuits, instrumentation, PCB design, microcontrollers, sensors, actuators, motor control, and embedded Rust.
Software & systems
Rust, Python, Linux, Docker, networking, data acquisition, automation, and systems programming.
Edge & accelerated computing
NVIDIA Jetson, CUDA, PyTorch, TensorRT, inference engines, GPU performance measurement, model optimization, and local AI.
Machines
Motion control, CNC and fabrication systems, robotics, machine vision, autonomous equipment, and other systems where electronics, mechanics, software, and computation have to work together.
Where This Is Going
gradientdescent.dev is increasingly focused on building complete physical systems.
The progression starts close to the fundamentals - circuits, components, measurement, firmware, and interfaces - and moves toward systems that combine sensing, computation, control, motion, and eventually intelligence at the edge.
Some projects will use AI. Many will not.
The common thread is learning how real engineered systems work from the component level upward, then using that understanding to design, build, instrument, and improve increasingly complex machines.
Everything begins on the bench.