This Machine Design article reports on Leo AI, a Cambridge-based startup whose “Large Mechanical Model” (LMM) is tailored for mechanical-design workflows rather than generic chatbots.
Key features:
- Leo AI is trained on more than one million vetted engineering sources plus customer-specific CAD/model data, and claims 96% accuracy versus ~46% for off-the-shelf large language models.
- It is geometry-aware: it can process B-rep CAD data, assemblies, and part geometry, not just textual descriptions. This lets it check design intent, search for parts, validate features, link to documentation, and surface organizational knowledge locked in PDM/PLM systems.
- The tool integrates smoothly into existing design toolchains and corporate data stores, enabling reuse of old work, improved access to engineering knowledge, and fewer lost hours searching or clarifying data.
From a workflow standpoint, mechanical-design teams stand to gain: routine tasks such as parts search, compliance checks, documentation drafts, and geometry review can be handed off to the AI-copilot. Meanwhile, engineers focus on deep design decisions. The article quotes Leo AI’s CEO saying that the issue isn’t that CAD tools are bad, but that organizational knowledge remains siloed in legacy files and silos.
CAD tool innovation isn’t just about generative geometries or new CAE features; it’s also about embedding intelligence that understands geometry, metadata, and enterprise-scale information. The result: improved throughput, less re-work, better knowledge capture, and faster decision-making in mechanical design.
Leo AI marks a shift from “AI for documents” to “AI for geometry and design data,” and engineers should pay attention.