Maya HTT has launched M-Bot, a locally deployable engineering knowledge assistant that searches technical documents, product lifecycle management records and industrial repositories. Using retrieval-augmented generation, it creates answers linked to the relevant source material while allowing companies to keep the knowledge base and processing within customer-controlled infrastructure.
Engineering organisations often hold design decisions across several generations of files and systems. Test reports may be stored alongside numerical results, while drawings, simulation outputs and manufacturing procedures are managed in separate repositories. A technician examining a recurring problem can therefore spend considerable time identifying which design revision applied, which assumptions governed a calculation and where the final qualification evidence was recorded.
Engineers can query internal records through a natural-language interface, drawing on Siemens Teamcenter and other product lifecycle management systems as well as repositories such as SharePoint and Confluence. The assistant is intended to find previous test results, manufacturing procedures, engineering calculations and information about known equipment problems, presenting the supporting records alongside its response.
M-Bot uses retrieval-augmented generation to identify relevant information in the company’s indexed repositories before preparing an answer. The links back to the original records give engineers a means to check the cited calculation, assumption or drawing revision against the question being investigated. Retrieval and answer generation remain dependent on the documents available and the quality of their indexing.
Technical files combine data formats that ordinary text search handles unevenly: numerical tables, X-Y plots, equations, design assumptions and simulation images. Maya HTT specifies support for contextual interpretation of finite element, computational fluid dynamics, thermal, electromagnetic and structural results, alongside the boundary conditions and governing equations associated with them. These capabilities are intended to make technical evidence searchable across different document formats.
A qualification report may record a peak temperature that applies only to a particular load, ambient condition or product revision. Reusing that value in a different design requires the associated test setup and acceptance criterion, not merely the number. M-Bot’s source links allow an engineer to retrieve the supporting plots and records before deciding whether the result applies to the present design.
Maya HTT offers on-premises servers, dedicated GPU hardware and customer-controlled private cloud deployments. It also describes configurations compatible with air-gapped networks, making the product suitable for investigation in environments where proprietary or export-controlled engineering information must remain under local governance. The company lists aerospace, defence and manufacturing among the target industries.
Role-based access control is part of the system’s design. A search across engineering repositories must respect programme restrictions and document permissions, particularly when projects contain material owned by different customers or governed by different export rules. During implementation, search permissions need to follow the organisation’s established access decisions as documents are added, revised or withdrawn.
The knowledge base and its embeddings — numerical representations used to retrieve related material — remain under the customer’s control, according to the supplier. The organisation can therefore manage the index alongside the original repositories, including retention, backup and access policies. Keeping data local also leaves the system dependent on customer-managed computing capacity and software maintenance.
Versioned records require particular attention. A retired procedure and its replacement may both contain the same technical vocabulary, and a previous drawing can appear relevant even after a design change. The assistant’s source references help an engineer distinguish between them when used with document status and approval information already maintained by the product lifecycle management system.
Maya HTT says M-Bot can retrieve technical information up to 60 times faster, illustrating the claim with a five-minute manual search reduced to about five seconds. Retrieval time across a deployed installation will depend on the repositories indexed, the local computing environment and the complexity of the question.
The company says its deployment model avoids fluctuating language-model token charges. With processing hosted on customer-controlled infrastructure, project costs instead include the computing hardware, configuration, repository connections, ongoing maintenance and support arrangements chosen for the installation.
Potential applications include design reviews, investigation of repeated faults, support for maintenance troubleshooting and transfer of specialist knowledge between experienced and newly recruited engineers. Each requires a different standard of evidence. A maintenance query may need the correct equipment configuration and current operating limits, while a design decision may require traceable calculations and formal technical sign-off before a proposed change can be approved.
M-Bot is available for demonstrations and enterprise deployment. Its integration with technical repositories and source-linked answers gives engineering teams another route to established design knowledge, while document control, technical review and formal sign-off remain embedded in the organisation’s normal workflow.
