What Ballista Labs is building
Ballista Labs builds Orville, an AI hardware-design agent that asks users to describe what a part should do, what it must fit and how it should be manufactured. Its current product page says Orville produces mathematically defined B-rep solids available in STEP format, with manufacturing-aware workflows for sheet cutting/bending, CNC machining and 3D printing. An API is also offered for integrating hardware generation into other agent workflows.
Where it sits in the engineering workflow
The emphasis on function and manufacturing process is important. 'Make a bracket' is not an engineering requirement. A useful mechanical-design agent needs envelope constraints, interfaces, fasteners, loads, process limits and material assumptions. Orville's product framing moves in that direction by asking users to include fit and manufacturing intent before generation.
Why mechanical engineers should care
For fixtures, brackets, enclosures and tooling, rapid CAD generation could reduce the drafting queue and make simple hardware work more accessible. The company's use of STEP/B-rep output also makes results easier to inspect in conventional CAD and manufacturing software than a visual mesh would be.
What it does not remove from engineering
Manufacturing-aware does not mean manufacturing-approved. A STEP solid does not contain all drawing requirements, tolerances, material certifications, finishes or inspection criteria. Structural or thermal screening results also need independent verification when they influence safety or compliance. The right professional workflow is generate, inspect, analyze, DFM-check and release—not prompt, export and manufacture blindly.
Questions to ask before adopting it
- Which manufacturing constraints are hard rules versus model suggestions?
- How does Orville handle missing loads, tolerances and interface requirements?
- Can a professional engineer continue editing the model parametrically after STEP export?
- What verification is expected before generated hardware is manufactured or used?
DIVE-LD engineering view
The most useful way to evaluate Orville is to put it inside a controlled engineering release process. Ask whether the output preserves design intent, whether assumptions are visible, whether another engineer can edit or audit the result, and what verification remains before manufacturing. AI can compress repetitive work and widen the design space, but responsibility for requirements, safety, standards, tolerances and final release still belongs to the engineering organization.
Sources and further reading
Public product information reviewed on 11 September 2026. Company statements about performance or capability are vendor claims unless independently verified.