Articles · November 12, 2026 · Luis Miguel Quevedo
From an idea to FPGA: where AI, MBSE and SysML v2 fit
A hardware idea often begins as an incomplete sentence. We need to detect something, control something, or move a task away from a larger computer. The problem is not that the sentence is imprecise. The problem begins when we start designing before turning it into engineering decisions.
I spent years working on systems where hardware was only one part of something larger. Robots, PLC control panels and industrial plants force you to think about interfaces, sequences, failures and constraints before choosing components. Today I work much closer to silicon, but the principle has not changed: writing RTL too early can mean solving the wrong problem with great precision. That industrial path and the transition toward digital design are what I now describe as moving from robots to silicon.
This is where MBSE becomes useful. Model-Based Systems Engineering does not mean drawing diagrams for their own sake. For me, it means using a model as the place where the need, requirements, architecture and verification can maintain explicit relationships. SysML v2 is particularly interesting because OMG's formal 2.0 specification includes textual syntax, defined semantics and a standardized API in addition to graphical representations. Those characteristics create useful opportunities for automation in hardware workflows.
AI can enter this process, but it should not replace engineering. A language model can help interpret an initial description, propose a first requirement structure or generate artifacts for later review. The important part is traceability. If a requirement changes, I want to know which architectural element it affects, which design depends on it and which verification needs to run again.
That was the central idea behind work we developed on AI-driven digital-design prototyping with MBSE for small-scale makers. I presented it at IEEE LAEDC 2025 in Guadalajara, where it received the Best Paper award.
The workflow I am building around this idea can be summarized simply. A person states the need. The system moves that need toward requirements and architecture. Hardware and firmware are designed against that structure. Simulation and FPGA or development-board prototyping follow. If the product justifies custom silicon, the process can continue toward physical implementation, manufacturing and integration. That same eight-stage sequence is currently published as the QvIC Labs cycle.
Not every product needs a chip. Not every project needs MBSE either. A simple board using well-known interfaces may be better served by a much more direct process. The value increases when dependencies grow, several disciplines need to coordinate, or an early architectural decision affects firmware, PCB, FPGA and silicon at the same time.
That is why I do not see AI, MBSE, SysML v2 and FPGA as four independent subjects. I see them as layers of the same engineering process. AI can reduce repetitive work. MBSE organizes decisions. SysML v2 can represent and connect the system. FPGA gives us a way to learn from an architecture before committing to a more expensive implementation.
The hard part remains human: defining the problem correctly and deciding what evidence will be sufficient to say that the system works.