Skip to the stories
Mechanical & Manufacturing

Xometry’s AI and CAD Integration Pushes Manufacturing Decisions Earlier

  • DEVELOP3D
  • 726 reads

Xometry is bringing process recommendations, pricing and manufacturability checks into design tools while applying AI to inspection and supplier workflows.

Xometry’s AI and CAD Integration Pushes Manufacturing Decisions Earlier

Xometry’s strategy is to bring manufacturing choices closer to the moment a part is designed. Its work on AI-based recommendations and integration with Siemens Designcenter reflects a broader shift: engineers want to assess manufacturability, cost and lead time before settling on a design iteration.

CTO Vaidy Raghavan described three areas of development over the past year and a half: reaching more purchasing channels and enterprise customers, expanding the company’s models, and using its manufacturing network to respond to demand across processes. Xometry connects product designers and engineers with manufacturing suppliers, and Raghavan said the company receives millions of part uploads each year. That data helps its models learn about both geometry and the intent behind a request.

Vaidy Raghavan, CTO, Xometry

Manufacturing feedback during design

Xometry has introduced a large language model based recommendation system for manufacturing processes, evolving an earlier version with additional data extraction. Raghavan said the company’s design for manufacturing logic has improved in accuracy. When a customer uploads a part, the system can flag potential risks and provide a price point and lead time options. He said a growing share of customers choose the recommended process, which he attributes to trust in those recommendations.

The Siemens Designcenter integration is intended to put these checks inside the design workflow. Raghavan framed the move as a response to “shift left” practices: bringing cost, lead time and manufacturability questions earlier, while designers can still make changes. In his view, waiting until a design is locked in makes those decisions harder to incorporate.

The integration is intended to expose more than a narrow selection of manufacturing options. Raghavan said Xometry’s catalogue of more than 2,000 materials and finishes will be available in Designcenter, allowing users to compare options and see pricing as they explore a design. That could make process selection a more iterative part of CAD work, with manufacturing constraints considered alongside geometry.

Xometry has offered plugins for several CAD packages for years, including Autodesk Fusion. For now, though, its focus is on getting the Designcenter experience and capability set right. Raghavan said the goal is to introduce new working habits for Designcenter users, not simply make the service accessible through another interface.

The delivery approach will account for both cloud and on-premises users. Siemens has customers in each group, and Xometry plans to address both while sequencing work to begin testing quickly. The company has not described a broader schedule for that rollout.

Scaling a global supplier network

Xometry reports rising demand across its manufacturing technologies, without one process dominating growth. That creates a capacity challenge: the supplier base needs to expand across sectors as orders increase. The company is also seeing demand for specialized certifications and requirements, which Raghavan cited alongside broader growth.

Xometry
Xometry offers custom manufacturing on demand to some 90,000 buyers worldwide through a network of more than 5,000 suppliers

Operating across regions brings differences in engineering documentation. Raghavan noted that manufacturers in European countries may expect drawings and CAD files in particular formats, while practices in the United States differ. A supplier network that works well locally cannot simply be expanded globally overnight. Xometry’s approach is to build on established regional ecosystems and bring suppliers into broader work gradually, working with them as they adapt.

That gradual expansion matters because the platform depends on suppliers being able to meet customer expectations across different markets. The task involves deciding which parts of a process can be standardized and where regional adjustments are needed. Raghavan described the growth of local networks and their gradual connection to a global system as a satisfying part of the work.

AI’s limits and quality applications

Raghavan said Xometry regularly tests emerging AI models for design for manufacturing risk assessment and price estimation. Their responses have become more relevant over the past year and a half, he said, but accuracy can fall sharply when part complexity increases. He also cautioned that fabricated model outputs remain a serious concern. For engineering decisions, a plausible answer is not enough; recommendations need to hold up against the part and the manufacturing process.

Other experiments use AI to match a part image with potential manufacturers by searching product catalogues. Xometry is also applying language and vision models to quality and documentation tasks where it says accuracy is already high enough to be useful. One example combines large language models with optical character recognition to detect handwriting and support automatically generated inspection reports.

The next area under consideration is live defect analysis from images captured during production. Raghavan described suppliers dividing a manufacturing sequence into stages, taking images between them and checking for quality problems before a part continues through the process. Catching a fault between stages can allow a defective part to be removed from the production cycle earlier.

With thousands of suppliers, Xometry can observe adoption across manufacturers at different stages and assess which capabilities fit particular operations. The company’s role, as Raghavan described it, includes connecting promising tools with suppliers for whom they make practical sense.

A more data-led production system

Raghavan sees AI moving manufacturing from a hardware-first model, built around plants, manual work and spreadsheets, toward one where software and data increasingly shape decisions. The potential advantage, in his account, comes from processing information quickly and using machine learning to support real-time choices.

His longer-term view is of a marketplace that learns continuously from manufacturing activity, sending information back to improve future decisions as well as informing current ones. Whether that model delivers depends on reliable data, useful recommendations and adoption across a varied supplier base. For engineers, the near-term change to watch is more immediate: manufacturing constraints and quality feedback appearing earlier in design and production workflows.

Source: DEVELOP3D · develop3d.com

From the same set