By VADM David Lewis (USN, Ret.), President of the American Society of Naval Engineers
These leading developments point toward engineering AI built on specialized models, native tools, and verifiable results, not general-purpose chatbots. Cadence introduced Maestro, a model portfolio for semiconductor and systems engineering. Nemetschek launched AI-enabled software covering structural modeling, analysis, detailing, and fabrication. PTC Onshape and SimScale demonstrated a workflow linking AI-generated parametric CAD functions to physics-based validation.
In maritime manufacturing, HII reported continued construction of its ROMULUS uncrewed vessels, while Fincantieri launched a workforce program linking shipbuilding with robotics, AI, and digital twins. Intel and SAP demonstrated enterprise software directing autonomous material handling, and an aerospace research partnership focused on adaptive robots for complex composite manufacturing.
These developments point to a common architecture:
Engineering definition → AI agent → validated engineering tools → manufacturing instructions → physical execution → inspection and feedback from operations.
More elements of this architecture now appear in commercial software and industrial demonstrations. Integrating them at scale remains the central challenge.
Key developments
1. HII expands distributed production of ROMULUS USVs
ROMULUS™ | HII
HII wins Navy contract for 10 ROMULUS uncrewed surface vessels | WorkBoat
Louisiana yard selected to build new USVs for naval use
HII marked ongoing construction of ROMULUS 151 uncrewed surface vessels at Breaux Brothers Enterprises in Louisiana by ceremonially installing a hull plaque on the first-in-class vessel. The milestone follows the Navy’s September 29 award for ten ROMULUS USVs. The production network includes Breaux Brothers, Halimar Shipyard, and Bayou Metal Supply & Manufacturing, with engineering support from Beier Integrated Systems and Incat Crowther. HII combines modular vessel architecture, distributed structural fabrication, and its Odyssey autonomous-control software.
Why it matters: ROMULUS tests whether modular design and digital manufacturing preparation can improve serial naval-vessel production. The approach could also support distributed construction of common structural assemblies, machinery modules, and other naval products.
This milestone confirms production activity, not productivity gains from HII’s HYPR robotics initiative. Key measures to track include labor hours per hull, structural rework, module completion rates, and integration time.
HII's October 28 production announcement
2. Fincantieri introduces an AI and robotics curriculum for shipbuilding
Fincantieri and ITS Academy Udine launched Mechatronics & Robotics – Smart Ship at Monfalcone. The two-year program provides 2,000 instructional hours, including 700 hours of company-based training, for an initial class of 16 students.
The curriculum covers industrial and collaborative robotics, AI, digital twins, production automation, predictive maintenance, onboard systems, and cybersecurity. It also teaches planning and production-management skills relevant to ship superintendents.
Why it matters: Fincantieri frames digital shipbuilding transformation as a workforce challenge, not just an equipment investment.
The program complements BOOST², Fincantieri’s industrial modernization initiative announced earlier in October. Combining technical education, hands-on production experience, and automation deployment could embed new manufacturing practices across the organization.
For American shipyards and engineering educators, the lesson is to teach CAD/CAM, robotics, digital engineering, manufacturing processes, and production management together rather than as separate specialties.
Fincantieri's Smart Ship program
3. Cadence Maestro brings specialized AI models to engineering agents
Cadence launched Maestro, a portfolio of engineering-specific AI models that works with its existing EDA and system-design automation software. Instead of relying solely on a general-purpose language model, Maestro combines domain-trained models, decision models, reusable agent skills, and reinforcement learning informed by established design and verification tools.
Cadence reports early workflow gains of up to 30% in design-result quality and fivefold greater token efficiency than general-purpose foundation models. The system supports cloud deployment and customer-controlled infrastructure, with select early access planned for the fourth quarter of 2026. These company-reported results do not establish universal productivity gains.
Why it matters: Semiconductor engineering offers a model for agentic systems engineering: an AI agent proposes changes, runs established tools, checks results against explicit constraints, and uses the evidence to guide its next decisions.
A similar approach in naval architecture could coordinate specialized agents for structures, arrangements, propulsion, electrical systems, producibility, and compliance against a shared design baseline. On-premises deployment is particularly relevant to defense engineering environments handling controlled technical information.
Cadence Maestro announcement
4. Nemetschek extends AI across structural modeling, analysis, and fabrication
Nemetschek introduced Allplan Max 2027, expanding AI across Allplan, SCIA, FRILO, RISA, and SDS2. Its tools support model creation, custom engineering functions, structural analysis, BIM interpretation, drawing automation, and fabrication-oriented design.
Companion products include Archicad Max, whose agentic assistant is in beta, and Bluebeam Max, which supports drawing comparisons, discrepancy detection, and automated markups. Early Allplan beta users reported completing selected tasks up to 36% faster. These self-reported results are not independently measured enterprise productivity gains.
Why it matters: The shared AI layer connects engineering disciplines, advancing the model-based engineering goal of keeping geometry, calculations, documentation, and construction information linked rather than passing disconnected artifacts between organizations.
In maritime engineering, a comparable ship product model could automatically identify the structural calculations, manufacturing details, drawings, and work packages affected by an approved design change.
Nemetschek's October 29 announcement
5. Onshape and SimScale link AI-generated parametric CAD with simulation
SimScale published a practical example of AI writing custom parametric modeling functions through Onshape’s FeatureScript interface, followed by a separate engineering agent evaluating the design through simulation. The workflow links a natural-language requirement to executable feature-generation code, an editable CAD model, and physics-based analysis. Demonstrated in recent technical sessions, the workflow received a new explanation on October 30—not a new product launch.
SimScale highlights a critical distinction: a plausible-looking 3D object may still lack the dimensional constraints, feature dependencies, and editable design history required for engineering.
Onshape’s October 30 update also added customer-configurable standard-component libraries, faster simulation through rigid-body simplification, and improved CAD-to-robotics export using simplified URDF kinematic structures. It also improved source traceability in AI Advisor.
Why it matters: The workflow brings together three requirements for trustworthy generative engineering: editable geometry, reusable engineering logic, and verification with established simulation tools.
Naval designers could apply this approach to standardized foundations, structural details, fluid-system components, and equipment mounting arrangements.
SimScale's October 30 technical walkthrough
6. A Small Industrial Engineering Team Shows Practical Uses For AI
A Siemens-sponsored engineering.com case study describes how GEHA, a Dutch manufacturer of large industrial fans, uses AI with Designcenter Solid Edge. Engineering manager Jan Bos uses AI to check drawings and bills of materials, flag manufacturing issues, write CAD automation scripts, and prepare production information. In one example, an AI-assisted review of approximately 150 drawings found inconsistent hole spacing and incomplete component specifications. Bos also uses AI coding tools to automate Solid Edge tasks and related manufacturing documentation.
Why it matters: The case illustrates useful AI applications without requiring a new CAD platform or an autonomous engineering organization. An experienced engineer can use commercial AI to capture recurring rules, check documentation, and reduce repetitive software tasks.
For smaller naval suppliers, AI could extend experienced engineers’ capacity before fully autonomous CAD becomes reliable. The boundary remains clear: AI assists with review and tool creation; engineers must assess correctness.
GEHA's AI-assisted engineering case study
7. Intel and SAP connect enterprise software to physical robots
Intel and SAP announced a proof of concept linking SAP Extended Warehouse Management directly to autonomous material-handling robots. At Intel’s Chengdu facility, enterprise material-movement requests become robotic tasks. The robots carry out those tasks and return completion details, status updates, and exceptions to SAP.
Intel says the architecture is independent of any single robotics vendor, operating system, AI model, or cloud deployment approach. The companies reported improvements in transport distance and repetitive handling but provided insufficient production baselines to establish broadly applicable savings.
Why it matters: Many manufacturers have deployed autonomous material vehicles without connecting them to enterprise production planning. This demonstration addresses that disconnect.
In a shipyard, this approach could link material-release decisions, work-package priorities, crane availability, and mobile robotic transport within a shared operational system. The potential closed loop is:
Production need → authorized material movement → robot execution → completion confirmation → production-state update.
Intel–SAP demonstration
8. Stereolabs adds industrial edge computing for multiview physical AI
Stereolabs introduced the ZED Box 2 industrial edge-computing series for autonomous mobile robots and other machine-vision applications. The platforms support up to eight synchronized cameras, combining multiple viewpoints with other sensor data. The higher-performance version uses NVIDIA Jetson Thor processing to run vision-language-action models and demanding perception workloads locally.
Why it matters: Adaptive manufacturing needs more than intelligent control: machines must also maintain an accurate enough representation of their surroundings.
In shipyards, multiview sensing could support automated inspection, collision avoidance, material handling, robotic assembly, and work inside incomplete structures. Local perception and reasoning reduce reliance on remote cloud processing, making industrial deployment more practical. This hardware enables automation; it does not demonstrate successful automation of a specific shipyard process.
9. Synopsys and NUS work toward physics-validated digital twins
Synopsys joined the National University of Singapore’s Sustainable Tropical Data Center Testbed Phase 2.0 consortium. The project combines high-fidelity computational fluid dynamics, multiphysics simulation, AI-assisted engineering, and reduced-order modeling with physical testbed measurements. Initial research focuses on advanced liquid cooling and predictive thermal-management digital twins. The partners plan to validate simulations experimentally before building faster models for operational decisions.
Why it matters: The project illustrates how physics-based simulation can produce computationally efficient models for operational use.
For naval applications, an engineering digital twin could predict equipment or system behavior quickly enough to inform operational decisions while remaining grounded in established physics and experimental evidence. Potential uses include shipboard thermal management, machinery cooling, ventilation, propulsion auxiliaries, and energy systems. The initiative remains research and validation work; it has not demonstrated operational energy savings.
Research highlights
Two additional research publications merit attention.
-
DIP-CAD, revised October 27, tackles a core generative-CAD challenge: preserving engineering relationships when dimensions or parameters change. It uses explicit geometric references and sketch constraints to retain design intent. Researchers report better geometric accuracy in initial models and after parameter edits—a step toward editable engineering models rather than static generated shapes.
-
An AI-driven warehouse digital twin paper published October 28 combines reinforcement learning, FlexSim simulation, and the Asset Administration Shell interoperability framework. Simulated warehouse experiments report a 29.66% reduction in average picking time, 35.42% higher throughput, and 18.54% less travel distance. These are simulation findings, not verified physical-factory outcomes. The approach is relevant to shipyard logistics, storage, staging, and material flow.
Conclusion
For maritime engineering and shipbuilding, the emerging opportunity is to connect editable design models, validated engineering tools, manufacturing information, and physical operations within a shared digital workflow. Near-term applications include documentation checks, reusable CAD automation, design-change coordination, and material handling—not fully autonomous ship design or production.
Realizing that opportunity requires workforce development, controlled technical data, and clear engineering accountability. Shipyards and suppliers should test bounded workflows and measure rework, labor hours, integration time, and material flow before scaling. Commercial announcements and simulation results show promise; sustained production evidence must establish the gains.
Main image caption: A U.S Navy civilian contractor tracks an Orca unmanned aerial system (UAS) for flight aboard the Trials and Research Vessel NAWC38 during the annual Fleet Experimentation (FLEX) 2026 event in Key West, Fla., April 27, 2026. Hosted by U.S. Naval Forces Southern Command/U.S. 4th Fleet April 24-30, the exercise integrates commercially developed unmanned systems and artificial intelligence with traditional manned naval platforms. FLEX 2026 features comprehensive collaboration across the Department of War and industry to demonstrate a sophisticated kill chain that successfully found, tracked and engaged captured drug boats. The campaign focuses on operationalizing advanced robotic and autonomous systems to combat transnational organized crime and patrol vast maritime regions. (U.S. Navy photo by Mass Communication Specialist 3rd Class Jasmin L. Aquino)