Axial3D and Compute Maritime: Why Niche AI Wins Where General AI Can't Compete

Company Profiles

Axial3D is a Belfast-based medical AI company led by CEO Roger Johnston. The company converts 2D medical imaging data — CT scans, MRI scans — into accurate 3D anatomical models used for surgical planning, patient-specific implant design, and the manufacture of surgical guides and anatomical models via 3D printing. The problem Axial3D solves is specific: 2D scan data has always contained 3D information, but extracting it accurately required specialist radiologists using manual segmentation tools, which was time-consuming, expensive, and inconsistent.

Compute Maritime is led by CEO Shahroz Khan and applies AI to naval architecture and maritime engineering — a field where the engineering complexity is extreme and the market size is small enough that general engineering AI companies have not built specialized solutions. Naval vessel design involves hydrodynamics, structural analysis, stability analysis, systems integration, regulatory compliance with class society rules, and eventually shipyard production planning — all of which are highly specialized and interact in complex ways.

Both companies have in common the strategic choice to go deep in a specific domain rather than broad across many engineering sectors. That choice is the primary driver of their competitive position.

---

The Challenge

Medical 3D Anatomy: Specialist Bottleneck at Scale

Surgeons preparing for complex procedures — orthopedic reconstruction, maxillofacial surgery, cardiac intervention, spinal surgery — increasingly want 3D anatomical context rather than 2D scan slices. 3D models allow surgeons to plan entry angles, identify anatomical variations, verify implant sizing, and discuss the procedure with patients more effectively.

The traditional path: a radiologist or medical engineer manually segments the CT or MRI data — tracing anatomical structures slice by slice in 3D editing software — and exports a 3D model. For a complex anatomy, this takes 4–8 hours of specialist time. At $150–$300 per hour for a specialist with the relevant skills, each case costs $600–$2,400 before printing. Delivery times of 2–5 days mean this is typically reserved for the most complex cases, not routine use.

This bottleneck limits adoption: surgeons who would benefit from 3D planning can't justify the cost and turnaround for every case. The specialist constraint prevents scaling.

Naval Architecture: Too Specialized for Horizontal AI

Naval vessel design does not fit into general engineering AI platforms because the domain is unique in almost every dimension:

Hydrodynamics: A vessel's resistance, seakeeping, and maneuverability require specialized analysis (towing tank testing or CFD specific to hull forms) that general fluid dynamics tools don't address well.

Regulatory framework: Every vessel that operates commercially must be classed by a recognized classification society (Lloyd's Register, Bureau Veritas, DNV, etc.). Class rules are extensive, technical, and change with new safety standards. Compliance analysis requires deep knowledge of these rules.

Structural analysis in a marine environment: Marine structures experience fatigue loading from wave-induced vibration and corrosive environments that are not well-represented in general FEA practice.

Shipyard production: Translating a vessel design into a build plan for a specific shipyard — with its particular berth size, crane capacity, panel line, and workforce skills — is a specialized production planning problem.

No general engineering AI platform has training data or domain models for these problems. The market is too small (there are far fewer shipyards than automotive factories) to justify the investment. Compute Maritime wins by being the only serious AI option for a market that needs AI but can't use what everyone else is building.

---

What Axial3D Built

Axial3D's core AI model was trained on a large library of annotated medical imaging data — CT scans with manually verified 3D segmentations across major anatomical structures. The training process captured the pattern-recognition expertise of the specialist radiologists who created the annotations.

The deployed workflow:

1. A surgeon or hospital uploads CT or MRI data through Axial3D's HIPAA-compliant platform
2. The AI automatically segments the relevant anatomical structures (bone, soft tissue, vasculature, organs — depending on the application)
3. A quality review step allows a clinical specialist to verify the segmentation before delivery
4. The verified 3D model is delivered as a DICOM 3D file, an STL file for 3D printing, or directly to the hospital's surgical planning software

Turnaround time: 4–8 hours for standard anatomies vs. 2–5 days for manual segmentation. For urgent cases, expedited service in under 2 hours.

Cost: 70–80% reduction compared to manual specialist segmentation services for standard anatomies.

Quality: Axial3D's published accuracy data shows segmentation accuracy comparable to expert radiologists across validated anatomy types. The AI is not better than the best expert — it is as good as a highly experienced specialist, available at scale.

The medical 3D printing application extends the value: once a 3D model exists, it can be 3D printed as a physical anatomical model (for surgical rehearsal), a surgical guide (for precise implant placement), or used for patient-specific implant design. Axial3D's platform supports the full workflow from scan to physical model.

The global reach implication: specialist radiologists capable of manual 3D segmentation are concentrated in major medical centers in developed countries. Axial3D's AI makes the capability available to hospitals without specialist radiology departments — including hospitals in countries where medical specialists are scarce. This is genuine democratization of a capability that was previously access-limited.

---

What Compute Maritime Built

Compute Maritime's platform addresses multiple phases of the vessel design and build process:

Concept design: AI-assisted hull form generation and optimization for resistance and seakeeping. Naval architects define mission requirements (speed, range, payload, sea states) and the AI generates hull form candidates that are optimized for hydrodynamic performance.

Regulatory compliance checking: The platform maintains an up-to-date rule set from major classification societies and checks design submissions automatically, flagging non-compliances before submission to the class society. This eliminates a significant portion of the review-revise cycle that currently adds weeks to classification.

Production planning: Translating a vessel design into a build plan for a specific shipyard — cut lists, weld sequences, block assembly plans — is currently done manually by experienced shipbuilding engineers. Compute Maritime's AI assists with production planning optimization, reducing the time from design approval to production start.

The market: naval architects, shipyards, and naval (government/defense) buyers. The size is small in absolute terms but the projects are large — a single vessel can be a $100M+ program — and the engineering complexity is high enough that even modest efficiency gains justify significant technology investment.

---

Results

Axial3D outcomes:

• Segmentation turnaround: 4–8 hours vs. 2–5 days for manual (70–90% reduction in calendar time)
• Cost per case: 70–80% reduction vs. manual specialist service
• Availability: hospitals with no in-house specialist radiology capability can now access 3D surgical planning support
• Surgical outcome correlation: early data from Axial3D clinical partners suggests reduced OR time and fewer intraoperative complications for complex cases where 3D planning was used, though prospective outcome data is still being collected

Compute Maritime outcomes:

• Concept design iteration: naval architects report evaluating 5–10x more hull form candidates per design phase using AI-assisted generation
• Compliance review cycles: automated rule checking reduces the number of classification society review rounds from 3–5 to 1–2 for standard vessel types
• Production planning: early deployments show 20–30% reduction in production planning time for standard vessel types, with more complex custom vessels showing smaller gains

---

Lessons Learned

1. The best AI moat is domain depth, not platform breadth. Axial3D's trained model is the result of years of annotation work on medical imaging data. Compute Maritime's rule knowledge represents codified expertise in naval architecture. Neither moat can be replicated quickly by a general engineering AI company.

2. Access democratization is underrated as a value proposition. Axial3D's ability to deliver specialist-quality segmentation to hospitals that cannot afford or access specialists is a different kind of value than efficiency improvement. It expands the market rather than just making the existing market more efficient.

3. Niche AI wins when the niche is large enough to sustain a company but small enough to be ignored by horizontal platforms. Maritime and medical imaging both fit this profile: important enough markets to support specialized companies, small enough that horizontal engineering AI companies don't invest in domain depth.

4. Validation requirements define the deployment path. Medical AI in clinical workflows must meet regulatory standards (FDA, CE mark, country-specific medical device approvals) that add 1–3 years to the path from prototype to market. Axial3D's regulatory strategy is as important as its technical development.

5. Human review preserves trust during AI rollout. Axial3D's quality review step — where a clinical specialist reviews AI-generated segmentations before delivery — is not a temporary crutch. It is the trust-building mechanism that allows AI to be used in clinical contexts where the cost of error is patient safety.

---

Implementation Advice

For hospitals and surgical centers: the Axial3D entry point is typically complex orthopedic or craniomaxillofacial cases where 3D planning has established clinical evidence. Start with the case types where the clinical value is clearest and the surgeon champions are most enthusiastic. Broaden from there as the workflow establishes itself.

For naval operators and shipyards: Compute Maritime's highest-ROI entry point is the regulatory compliance workflow — automated rule checking pays for itself quickly by reducing the review cycle, and it does not require changing the core design workflow. Add hydrodynamic optimization and production planning in subsequent phases.

---

About the Source

This case study is drawn from AI Across the Product Lifecycle Episode 24, a podcast conversation with Roger Johnston (CEO, Axial3D) and Shahroz Khan (CEO, Compute Maritime). See also: [[Medical Device PLM]], [[AI in Manufacturing]], [[Simulation and PLM]], [[Digital Twin in Manufacturing]].