Architecture for flight, orbit, motion and the body.
The instrument is one thing. What changes between sectors is the physics it has to respect, the qualification it has to survive, and the cost of being wrong. We work where the constraint count is high and intuition fails earliest.
Aerospace
Structural and thermal components where mass is a recurring cost for the life of the airframe, and where every unfamiliar geometry has to survive a certification conversation.
Typical problems: brackets and fittings at minimum mass under multi-axis load; heat exchangers and cold plates with a fixed envelope; secondary structure where the assembly count is the real cost driver.
Generative structures — spacecraft antennas, aircraft partitions and engine brackets — have already produced forms that outperform their human-designed predecessors on weight, strength, or both.
Defence
Mission-critical parts under hard qualification requirements, constrained supply, and a design review that will ask why the geometry looks like that.
Typical problems: sustainment parts where the original supplier is gone; weight and signature trades held simultaneously; components that must be producible at a forward location rather than only at a plant.
Where a supply chain is the binding constraint, the search has to include what can actually be made, not only what would perform.
Space
Launch and orbital structures where a prototype cannot be recalled, mass is the budget, and thermal cycling does things to a part that a static analysis will not show you.
Typical problems: primary and secondary structure at minimum mass; propulsion components with internal geometry no subtractive process can produce; thermal paths through a structure that also has to carry load.
Components designed under this approach have flown — including on lunar and Mars missions.
Medical
Devices and implants that have to satisfy physics, regulation, and a frightened patient's hand. The optimiser does not know the last of those, which is exactly why the brief matters.
Typical problems: patient-specific implant geometry with bone-matched stiffness; instrument design where ergonomics is a hard constraint rather than a finishing touch; lattice structures tuned for osseointegration.
Personalised treatment turns on dozens of patient-specific variables — genome, history, comorbidities, interactions. No clinician can hold them all; a model can. What comes back is a larger menu, not a decision.
Automotive
Structure, thermal and packaging problems where twenty objectives compete, none can be dropped, and the volume makes a gram worth arguing about.
Typical problems: battery enclosure and thermal management under crash load; casting and node consolidation; tooling and fixtures produced faster than a supplier can quote them.
Robotics
Actuated systems where geometry, control and materials are one object. Designing them separately is how you get a machine that works on paper and oscillates in the room.
Typical problems: limb and linkage design at minimum inertia; compliant mechanisms designed as a single part; end effectors tuned to a specific grasp rather than a general one.
What the sectors share.
High constraint count
A mind holds four to seven concepts at once; these problems carry twenty interacting variables. A person has to prune before they can start, and every prune closes a door on a region nobody has looked at.
Expensive prototypes
When building a candidate costs weeks or a launch slot, the value of evaluating it in simulation rises by the same factor. This is where the cost of exploration collapses.
A real review
Somebody has to sign. An unfamiliar shape with no argument behind it does not pass, and should not — which is why the reasoning ships with the geometry.
If your sector is not listed, the question is whether the physics is modellable — not whether we have done it before.