About us
One engineer's idea,
built in the open
Engineering programs fail at the seams: between the SysML model and the spec, between the spec and the test, between the test and the evidence. Requirements drift. Traceability breaks. By the time anyone notices, months of design work are already committed.
An ontology-grounded engineering graph came first, and building it is where the useful lesson came from: a requirement is only worth having if something can check it, and a finding that does not carry a number cannot be acted on.
Vector Stream Systems builds MBEstudio to close those gaps. It connects SysML v2 source, geometry, interfaces, budgets and requirements in one workspace. State simulation evaluates requirements as model values change, and declared trade studies compare candidate designs against their objectives and constraints.
The same conviction drives the other half of the work: a systems model should be something you can stand up in an afternoon, not a quarter. Sketch the assembly, state the budgets as constraints rather than prose, and get a model you can argue with the same day. Rapid prototyping and rigor are not opposites: the model earns its rigor by being cheap enough to build early.
The model stays yours. Every finding carries a number.
Rapid prototyping
A systems model should cost hours, not quarters. Stand one up from the specs and drawings you already have, then throw it away and try the next architecture. Cheap models get built early, and early models are the ones that change decisions.
Provenance by default
Every number is derived, not retyped. Every constraint is executable. Every finding says what it measured and what it measured against.
Structural clarity
Dependencies should be inspectable, not guessed from slide decks. We make the invisible visible.
Accountability
Humans stay in command. The tool measures and reports; people decide what the limits are and own the trade.
Behind the work
Built with patience, tested against the edge
Vector Stream Systems is a one-engineer shop building the tools I needed on real programs and never had. Everything here is written and maintained by one person, iterated in public on GitHub, and shipped when the code is right rather than when the calendar says so. That constraint is also the product thesis: if one engineer can prototype a full systems model in an afternoon, so can your team.
That engineer is Matthew Collaro. If you want to connect professionally, reach out on LinkedIn or get in touch directly.
Matthew Collaro
Founder & engineer, Vector Stream Systems LLC
Registered in Washington
Back the work
One machine short
A one-engineer shop has one real ceiling on what it can build, and mine is compute. How large a reasoning model the analysis pass can hold is set by the machine underneath it, which today means a cloud round-trip for anything serious, an awkward position for a tool whose whole argument is that your models stay yours.
An NVIDIA Jetson AGX Thor closes that in one box. Its 128 GB of unified memory holds a large reasoning model locally, and it is the same class of edge hardware that ends up inside the robots and vehicles these models describe, so the analysis pass can be developed and tested on the machine it would actually run on. It is the single piece of hardware that would most change what I can ship.
Jetson AGX Thor specifications
Shareable link: vectorstreamsystems.com/thor
Get in touch
Questions about MBEstudio, partnership inquiries, or just want to talk shop about systems engineering.