aerodAInamics

Origin

GMU capstone produced a full MBSE/SysML design blueprint and algorithm selection (Graph Neural Network); no code exists yet.

The Full Story

aerodAInamics started as a question inside MZI's own engineering ideation practice: could machine learning make aerodynamic analysis accessible to teams that cannot afford wind tunnels or heavy computational fluid dynamics work. Aircraft manufacturers, race teams, and wind turbine designers all face the same bottleneck: aerodynamic testing is expensive and slow, and that cost keeps smaller players out of the field.

The idea did not move forward right away. AI tooling capable of doing this kind of work well was not mature enough at the time, so it stayed a business case rather than a working product for a couple of years. MZI later picked it back up and sponsored a graduate capstone course with George Mason University's systems engineering program, and a student team took on the project.

Their work is real engineering, not just a slide deck. The team used SysML to define the system architecture, then ran a formal comparison across several machine learning approaches, including random forest, support vector machines, and convolutional neural networks, before recommending graph neural networks as the right fit for modeling airflow behavior.

No code exists yet. The capstone produced a design, not a working model, which is why aerodAInamics sits at the Spark stage: a real, well-scoped idea with serious engineering thinking behind it, but nothing built. Turning it into a working prototype is the logical next step, and it has not been scheduled yet.