Building the World’s First AI-Powered Custom Orthotics Manufacturing System
Service:
Custom AI Development, Computer Vision, Full-Stack Mobile & Web Development
Client:
DocSols
Industry:
Healthcare Technology, Medical Devices, Podiatry
Location:
Australia, Asia-Pacific
Client Website:
Challenges
Why Is Manual Podiatric Assessment Impossible to Scale?
Every foot is different. That sentence sounds simple. But for a company trying to manufacture custom orthotics at scale, it is the single most brutal constraint in the entire business model.
Traditional podiatric assessment relies on a trained clinician sitting across from a patient, interpreting scan data, making judgment calls built from years of clinical experience, and then translating all of that into a prescription. That process is slow. It is expensive. And it cannot be replicated across a distributed network of clinics without hiring more podiatrists, which compounds costs faster than revenue can keep up.
DocSols understood this. They were not trying to replace podiatrists. They were trying to give each podiatrist the leverage of three, without burning them out or cutting corners on patient outcomes. The gap between that ambition and what existing software could offer them was enormous.
How Do You Convert Physical Foot Scans into Manufacturing-Ready STL Files Automatically?
The jump from a digital foot scan to a 3D-printable STL file is not a small one. It involves interpreting complex spatial data, identifying structural anomalies, cross-referencing biomechanical load patterns, and generating a prescription that a manufacturing system can act on directly.
Before this project, that entire chain required human intervention at multiple points. A clinician would capture the scan. Another review step would diagnose it. A separate workflow would generate the prescription. Then a technician would model the orthotic. The number of handoff points was high. Each one introduced delay. Each one introduced the risk of inconsistency.
DocSols needed a system where a clinician could initiate a scan and walk away while software handled everything downstream. That system did not exist.
What Does It Take to Build a Patentable AI System from the Ground Up?
Building production-grade computer vision for medical applications is not a side project. It requires deep learning architecture designed specifically for the domain, training data pipelines that handle the noise and variability of real clinical scans, rigorous accuracy benchmarking against clinical standards, and an inference layer that is fast enough to be useful inside a clinical workflow.
DocSols needed a partner who could own this problem end to end. Not just the AI model in isolation, but the mobile capture experience that feeds it, the backend that processes it, the admin layer that manages it, and the manufacturing output it drives. The entire stack needed to be built, integrated, and delivered as one coherent product.
That is a different brief from most software projects. And it demanded a different kind of partner.
Solution
DSet: Proprietary Deep Learning for Foot Analysis
The centrepiece of everything PrimeSens built is DSet, short for Doc Sols Eye Technology. It is a proprietary artificial intelligence engine built on deep learning models trained to interpret foot scan data with clinical-grade precision.
DSet performs four core functions inside the orthotics manufacturing workflow:
Triage: Classifying incoming scan data and flagging cases that require immediate clinical attention versus those suitable for automated processing
Progressive treatment mapping: Identifying biomechanical patterns and matching them to evidence-based treatment pathways
Proposal prescription generation: Producing a structured orthotic prescription without human input, ready for clinician review or direct manufacturing
STL file generation: Translating the prescription into a 3D-printable file formatted for direct integration with manufacturing hardware
DSet reaches an accuracy level of up to 96%. That number was not handed to us. It was engineered through iterative model training, clinical feedback loops, and continuous benchmarking against real podiatric assessments. The system is currently undergoing patent registration, which reflects both the novelty of the approach and the depth of the intellectual property created.
Full-Stack Platform Built for Clinical Environments
AI alone does not transform a clinical workflow. The surrounding software has to work just as hard. PrimeSens built the complete software ecosystem that DocSols needed to put DSet into the hands of clinicians across Australia and the Asia-Pacific region.
Android and iOS Mobile Applications
Clinicians use the mobile apps to initiate and capture digital foot scans. The interface was designed specifically for clinical settings, where speed, accuracy, and minimal friction matter. A podiatrist does not have time to navigate a complicated UI between patients. The capture experience is clean, fast, and purpose-built for the task.
The apps communicate directly with the DSet backend, passing scan data upstream for AI processing and returning results in a format the clinician can read, review, and act on within the same session.
Backend Infrastructure
The backend handles the computational heavy lifting. Scan data comes in, gets processed through DSet, and outputs flow back out to the right places, whether that is the clinician's mobile device, the admin dashboard, or the manufacturing pipeline. The architecture was built to handle concurrent clinical sessions across multiple locations without degrading performance.
Security and data integrity were non-negotiable. Patient scan data carries clinical sensitivity. The backend infrastructure was designed with appropriate access controls, encryption standards, and audit trails to meet the expectations of a healthcare technology company operating across multiple markets.
Admin Dashboard
DocSols needed visibility across the entire operation. The admin dashboard gives the DocSols team a real-time view of cases moving through the system, prescription outputs, manufacturing queue status, and AI performance metrics. It is the operational nerve centre that ties every piece of the platform together.
Managing a distributed network of clinics without this kind of centralised visibility would mean flying blind. The dashboard removes that problem entirely.
Semi-Automating the Podiatrist's Role
The most significant design decision in this project was the concept of semi-automation. DSet does not attempt to replace podiatric expertise. It absorbs the repeatable, interpretable parts of the assessment workflow, triage, pattern recognition, prescription formatting, STL generation, and hands the clinician a pre-processed recommendation rather than a raw data dump.
A podiatrist working with DSet spends their cognitive energy on review, refinement, and patient interaction rather than on the mechanical interpretation steps that the AI handles faster and more consistently. That distinction matters. It is what makes the system clinically credible rather than just technically impressive.
The result is a podiatrist who can see more patients, deliver more consistent outcomes, and spend their expertise where it actually creates value. DocSols made their podiatrists 3 times more efficient. That is not a product marketing claim. That is the operational outcome of a well-designed semi-automation architecture.
Results
3x podiatrist efficiency gain
96% AI diagnostic accuracy achieved
70% reduction in prescription turnaround time
Full digital scan-to-STL pipeline delivered with zero manual modelling steps required
Platform successfully deployed across clinical environments in Australia and the Asia-Pacific region
DSet computer vision engine entered patent registration, representing protectable intellectual property for DocSols
Mobile capture applications live on both Android and iOS, with clinical-grade UX validated in active podiatry settings
Admin dashboard providing real-time operational visibility across distributed clinic network
Manufacturing output quality improved through consistent, AI-generated STL file specifications
DocSols positioned as the operator of the world's first AI-based custom foot orthotics manufacturing system
If you are building a product that sits at the intersection of AI, healthcare, or specialised industry automation and you want a partner who can own the entire stack from model to mobile app, PrimeSens is the team to call. We do not just consult on strategy. We build the thing.
Frequently Asked Questions
What is AI-based custom orthotics manufacturing?
AI-based custom orthotics manufacturing uses artificial intelligence to interpret digital foot scan data and automatically generate orthotic prescriptions and 3D-printable design files. The process replaces manual clinical interpretation steps with deep learning models trained on biomechanical data. The result is faster, more consistent orthotic production without sacrificing clinical accuracy.
How accurate can AI be in medical device design?
AI systems built for specific medical domains can reach high accuracy levels when trained on quality domain data and validated against clinical benchmarks. DSet, the computer vision engine built for DocSols, reaches an accuracy level of up to 96% in foot analysis and prescription generation. Accuracy depends heavily on training data quality, model architecture, and the specificity of the clinical task being automated.
What is computer vision and how is it used in healthcare?
Computer vision is a branch of artificial intelligence that trains systems to interpret and analyse visual data, such as images and scans. In healthcare, computer vision is used to analyse medical imaging, detect anomalies, classify conditions, and generate structured outputs from visual inputs. In the context of podiatry and orthotics, computer vision interprets 3D foot scan data to produce clinical assessments and manufacturing specifications.
How long does it take to build a custom AI system for a healthcare application?
Timeline depends on the complexity of the AI task, the availability and quality of training data, the regulatory environment, and the scope of surrounding software infrastructure. A production-grade computer vision system built for a specific clinical use case typically takes between 6 and 18 months from scoping to deployment. Projects that also require full-stack platform development alongside the AI engine sit toward the longer end of that range.
What technology stack is typically used to build AI-powered healthcare platforms?
A production AI healthcare platform typically involves several layers:
AI and machine learning: Python, TensorFlow, PyTorch, or custom deep learning frameworks for model development
Mobile applications: React Native, Swift (iOS), or Kotlin (Android) for clinical capture interfaces
Backend infrastructure: Node.js, Python (FastAPI or Django), or Go for API and processing layers
Cloud infrastructure: AWS, Google Cloud, or Azure for scalable compute, storage, and security
Admin and ops tooling: React or Next.js for dashboard interfaces
Data pipelines: Apache Kafka, Airflow, or custom ETL systems for scan data processing
The right stack depends on the specific performance, security, and scalability requirements of the use case.
Can AI semi-automate a specialist's job without replacing them?
Yes. Semi-automation is a well-established design pattern in high-skill professional environments. The goal is not to remove the specialist but to remove the repeatable, interpretable steps that do not require specialist-level judgment. The specialist then applies their expertise to review, edge cases, and patient interaction, which is where their knowledge creates the most value. In the DocSols case, podiatrists shifted from manual scan interpretation to reviewing AI-generated prescriptions, which is a far better use of clinical expertise.
What is an STL file and why does it matter in orthotics manufacturing?
An STL file is a standard 3D model format used in additive manufacturing and 3D printing. In orthotics, an STL file contains the complete geometric specification of the orthotic device, including arch profile, heel cup depth, and structural contours. Manufacturing hardware reads the STL file directly and prints the orthotic to exact specification. Generating STL files automatically from AI prescriptions removes the need for a technician to manually model each orthotic, which is one of the key bottlenecks DSet eliminates.
How do businesses protect AI systems they commission?
AI systems can be protected through a combination of patent registration, trade secret protections, and contractual IP assignment. A custom AI model trained on proprietary data with a novel architecture is potentially patentable, as demonstrated by DSet entering the patent registration process. Businesses commissioning custom AI development should ensure their contracts clearly assign intellectual property rights to them, not the development partner.
What should a business look for in an AI development partner?
A capable AI development partner should demonstrate:
Proven experience building and deploying production AI systems, not just prototypes
The ability to own the full stack, from data pipelines to user interfaces, not just the model layer
Domain familiarity with your industry or the ability to acquire it quickly through structured discovery
Transparent methodology for model training, accuracy benchmarking, and validation
Clear IP assignment practices and contractual protections
Experience working within relevant regulatory or compliance frameworks, particularly in healthcare or finance
The difference between a partner who builds a model and a partner who delivers a working system is significant. Most businesses need the latter.
Is it worth investing in custom AI development versus off-the-shelf AI tools?
Off-the-shelf AI tools work well for generic tasks: content generation, basic data classification, customer support automation. They do not work for tasks that are domain-specific, clinically sensitive, or competitively differentiated. DocSols could not have built the world's first AI-based orthotics manufacturing system using a generic AI product. Custom AI development is the right investment when the task is specific enough that no existing tool can handle it, and when the output represents a genuine competitive advantage. In those situations, commissioning custom development is not a cost. It is the foundation of a defensible business.
How can AI automation improve throughput in healthcare clinics?
AI automation reduces the time clinicians spend on repeatable cognitive tasks, triage, interpretation, documentation, and prescription formatting. When these steps are handled by software, clinicians can move through more patient cases within the same working hours. In the DocSols implementation, podiatrists became 3 times more efficient without increasing headcount. For clinic operators, that translates directly into higher revenue per clinician, shorter patient wait times, and the ability to scale geographically without a proportional increase in specialist staffing.
