I spent six years designing pipe racks and equipment platforms for Gulf Coast oil and gas facilities.
Now I also build AI agent systems that run six businesses.
These two things are not as different as they sound.
Structural Engineering as a Mental Model
Structural engineering teaches you to think in systems. Every element has a load path. Every load path has a failure mode. You design for the failure mode before you design for the load.
AI agent architecture works the same way.
Every agent has an input path. Every input path has a failure mode — hallucination, timeout, bad routing, context loss. You design the guardrails before you design the workflow.
What Engineering Gave Me
- Tolerance for specification-heavy documentation
- Comfort with systems that have real consequences if they fail
- Habit of verifying assumptions before building
- Understanding that complexity compounds — keep it simple at the component level
All four apply directly to building production AI systems.
The Translation
| Engineering Concept | AI Equivalent |
|---|---|
| Load path | Data flow between agents |
| Failure mode | Hallucination or routing error |
| Code reference (AISC, ACI) | LLM system prompt constraints |
| Shop drawings | Agent tool schemas |
| RFI | Human approval gate |
| Punch list | Audit agent output |
The Path
I did not leave engineering to build software. I layered software on top of engineering.
The job funds the empire build. The empire builds the exit. The engineering knowledge makes both more defensible.
You do not need to choose between technical depth and building businesses. The technical depth is the competitive advantage.