The most valuable part a manufacturer can design is often one it already has, buried somewhere in a library of thousands of legacy CAD files. The problem is finding it. DeepC3, a new startup, is building a search engine for the physical world, one that treats a part’s geometry as its primary data point. Its product, CADDLE, is a geometric data warehouse designed to index every component a company has ever designed, making it searchable by shape, physical property, or a simple English description, all while running on a customer’s own servers [DeepC3, retrieved 2026].
For engineers in aerospace, defense, or any sector handling ITAR-controlled data, the promise of a powerful search tool that never sends a file to the cloud is a compelling one. DeepC3’s founding thesis, articulated in a June 2026 blog post, is that on-premise operation isn’t just a preference for these organizations, it’s a non-negotiable requirement [DeepC3, June 2026]. The company’s founder, an eight-year veteran of simulation engineering, built CADDLE to parse geometric and physical properties together, returning ranked matches without requiring engineers to maintain a perfect taxonomy of tags or metadata [DeepC3, June 2026].
The On-Premise Wedge
DeepC3’s initial wedge is clear: compliance and control. By designing CADDLE to run entirely on a customer’s hardware, the company directly addresses the security and regulatory hurdles that keep sensitive CAD libraries siloed and inefficient [DeepC3, June 2026]. The product is CAD-neutral and PLM-independent, aiming to sit across a company’s existing software stack rather than replace it. This architecture allows it to ingest a library, create a searchable embedding space from the geometry of each part, and enable queries like ‘flat bracket with mounting holes’ [DeepC3, retrieved 2026].
The immediate value proposition for potential design partners is tangible savings through part consolidation. In a demo, CADDLE indexed 103,138 parts and surfaced 5,505 duplicate clusters, highlighting an estimated $2.49 million in open consolidation savings [DeepC3, retrieved 2026]. The company claims shape similarity searches across this scale take just 42 milliseconds [DeepC3, retrieved 2026]. DeepC3 is currently seeking a small number of design partners, offering to analyze a sample library and return a report detailing these duplicate clusters and potential cost savings [DeepC3, June 2026].
An Honest Counterfactual
The bet is ambitious, and the path to commercial traction is untested. The public record shows no announced customers, funding rounds, or named partners beyond the company’s own invitation. The competitive landscape for CAD search and data management is also well-established, with large incumbents embedded in engineering workflows. DeepC3’s answer appears to be specificity and architecture. Its focus on the on-premise, compliance-sensitive segment of manufacturing is a deliberate niche. The technical differentiator rests on the proprietary method of generating and querying geometric embeddings locally, a claim that will need to be validated in real-world, multi-terabyte libraries beyond a controlled demo.
The standard workflow in heavy manufacturing today is often a frustrating blend of memory, manual folder navigation, and brittle part-numbering systems. Engineers waste time redesigning components that already exist because they cannot be found. For the patient population of mechanical engineers and procurement managers in regulated industries, the current standard of care is inefficient and expensive. DeepC3 is betting that the cost of that inefficiency, measured in redundant engineering hours and unnecessary inventory, is high enough to justify a new, specialized layer of geometric intelligence.
Sources
- [DeepC3, June 2026] ITAR, the Cloud, and Your CAD Library: Why On-Premise Matters | https://deepc3.com/blog/itar-the-cloud-and-your-cad-library.html
- [DeepC3, retrieved 2026] CADDLE, The Geometric Data Warehouse for Manufacturing | https://deepc3.com/
- [PERPLEXITY SONAR PRO BRIEF, retrieved 2026] PERPLEXITY SONAR PRO BRIEF