Medical Device Software Development: Six Stages and Seven Engineering Partners
- IoT.Business.News
- 849 reads
A practical guide to medical device software stages, project evidence and seven firms with experience in connected and regulated products.



Choosing a medical device software team is a project design decision. The work spans intended use, software architecture, risk controls, testing, submission materials and updates after launch. A vendor that can show how those pieces connect gives a device maker a clearer basis for comparing proposals and spotting missing work before it becomes expensive to fix.
Connected devices add another set of interfaces to manage. Wearables, mobile apps, cloud services and hospital systems can exchange data in real time, so teams must account for connectivity and secure communication alongside the device’s core function. The practical question is not simply whether a supplier can write the software; it’s whether the supplier can keep the technical work and its evidence aligned through the product’s life.
Six stages, with evidence at each handoff
Medical device software work is commonly organized into six stages. Each should produce records the device maker can review, and each leaves decisions with the manufacturer. The documents provide a way to check progress beyond demonstrations of working code.
1. Define intended use and classification
The first task is to describe who will use the device, which patients it serves and what clinical purpose it supports. That intended use informs the product’s pathway: in the US, this may involve 510(k), De Novo or PMA; in the EU, the device is assigned an MDR class. The software also receives an IEC 62304 safety class, A, B or C.
Expected outputs include a scope document, a written rationale for the classifications and an initial requirements list. Those records give the team a shared definition of what it is building before detailed architecture or implementation begins.
2. Design the architecture and plan for risk
Engineers divide the software into units and identify which ones could contribute to harm. They then plan controls under ISO 14971. Third-party software components, known in IEC 62304 as SOUP, also need to be inventoried and assessed.
Architecture work covers the device’s connections to apps, cloud services and hospital systems such as electronic health records. The stage should leave an architecture document, a risk management plan and a traceability matrix linking requirements to design elements and tests. With those in hand, the manufacturer can decide whether the proposed scope and risk controls justify funding the build.
3. Build in iterations and keep records current
Short development cycles let the product team inspect working software and adjust priorities as the project progresses. Unit tests, code reviews and risk records should move forward alongside each build. When the scope changes, the change needs the same risk review as the original requirements.
This approach keeps documentation connected to the software being delivered. It also gives the manufacturer a regular opportunity to decide whether a proposed change is worth including before the next iteration starts.
4. Verify requirements and validate use
Verification asks whether the software meets its written requirements. Validation asks whether the device meets user needs under real conditions. Testing should follow the requirements specification, with effort prioritized according to risk; usability work is guided by IEC 62366-1.
Deliverables include test protocols, results and traceability records that connect each requirement to evidence. Clinical and quality leads at the manufacturer decide whether that evidence is ready to support a filing.
5. Prepare software materials for submission
The submission package draws on records from earlier stages: requirements, architecture, risk files and test results. For connected devices, it also includes a cybersecurity plan and a software bill of materials, or SBOM. FDA guidance sets Basic and Enhanced documentation levels for device software functions; the risk posed by the software determines which level applies.
A development partner may prepare the software sections of the technical file and respond to questions about them. The manufacturer remains responsible for filing and for the device associated with that filing.
6. Manage release and later changes
Release begins an ongoing cycle of change assessment, regression testing and record updates. Patches, new features and security fixes all need review, and some changes can require another submission. Complaints, device logs and telemetry can feed field information back into the risk file.
The vendor can prepare an impact analysis, but the manufacturer decides which changes to ship and when. That decision depends on how the update affects the device, its documented risks and the evidence supporting it.
Seven firms with medical device software experience
The companies below describe work across software as a medical device (SaMD), embedded systems, connected products and wearables. Their reported certifications, service areas and project examples can help narrow a shortlist. They don’t replace a project-specific review of staffing, deliverables, ownership terms and experience with the device’s intended use.
| Company | Founded | Clutch rating | Reported focus | Reported certifications or standards | Project examples |
|---|---|---|---|---|---|
| Relevant Software | 2013 | 4.9 (32 reviews) | Wearables, IoMT, AI, SaMD | ISO 27001, HIPAA, GDPR | Össur ordering apps; AstraZeneca analytics automating 35K+ records |
| ScienceSoft | 1989 | 4.8 (43 reviews) | SaMD, embedded and connected devices, mobile, AI | ISO 13485, ISO 27001, IEC 62304, HIPAA | CE-marked lab diagnostics software; imaging demo with 3,400+ test cases |
| Itransition | 1998 | 4.9 (42 reviews) | SaMD, in-hospital, in-home and on-body devices, IoMT | IEC 62304, FDA 21 CFR 820 and Part 11, HIPAA | Terumo connectivity software; asthma app with 5x lower change failure rate |
| Innowise | 2007 | 4.9 (73 reviews) | SaMD, embedded, IoMT, wearables, AI | ISO 13485, ISO 9001, ISO 27001, IEC 62304 | Smart ECG monitor Bluetooth Low Energy integration in 6 weeks |
| Softeq | 1997 | 4.9 (27 reviews) | Hardware, embedded, IoMT, remote patient monitoring, AI | ISO 13485, ISO 27001, ISO 9001 | Heart simulator with 0.1 ms timing accuracy; glucose prototype with 25% less prototyping time |
| Sigma Software | 2002 | 4.8 (37 reviews) | Device data and cloud analytics, remote monitoring analytics, embedded and IoT | ISO 13485, ISO 9001, ISO 27001 | Siemens Healthineers CT monitoring migration to Azure |
| SumatoSoft | 2012 | 4.8 (25 reviews) | IoMT, wearables, companion apps, AI | ISO 27001, ISO 9001 | Continuous glucose monitoring app; blood pressure cuff app with 20,000+ downloads |
Relevant Software
Founded in 2013, Relevant Software reports more than 200 delivered projects, with work spanning connected products, wearables and AI features. Its stated services cover early architecture, risk records and traceability; wearable and companion apps for Apple Watch, Wear OS and HealthKit; LTE-M connectivity, telemetry and AI for diagnostics and imaging; and updates after launch.
Reported examples include prosthetic and orthotic ordering apps for Össur that use 3D sensor scanning. The company says measurement and ordering time fell from hours to under 5 minutes. It also reports building an AI medical affairs analytics system for AstraZeneca that automated 35K+ medical records. Its profile lists ISO 27001, HIPAA and GDPR, and says 92% of its engineers are senior and employee retention is 96%. Claims of AI-assisted delivery, including cycles up to 50% faster and up to 20% fewer bugs, should be assessed against the supplier’s project evidence.
ScienceSoft
ScienceSoft, founded in 1989, reports a 750+ person IT workforce and work on Class II and III devices. Its scope includes SaMD, embedded and connected systems, mobile applications, AI models and cloud services on AWS, Azure and Google Cloud. The company describes development under ISO 13485, IEC 62304 and IEC 82304-1, with design history files.
Its listed project examples are CE-marked laboratory diagnostics software for a genetic screening kit maker, completed in 7 months, and a SaMD imaging web demonstration with 3,400+ manual test cases and reported 100% test coverage.
Itransition
Itransition has delivered software since 1998 and reports 3,000+ specialists and 1,530+ projects. Its device work covers in-hospital, in-home and on-body products, including experience with FDA Class II and III devices and IEC 62304 Classes A, B and C. Listed capabilities include SaMD, IoMT, mobile apps and cloud backends, with work mapped to IEC 62304, FDA 21 CFR Part 820 and Part 11, and HIPAA.
One example is TOMEs, which connects and monitors blood collection devices for Terumo Blood and Cell Technologies and is now in its seventh version. Itransition also reports an asthma management app for a respiratory device startup, with a 5x lower change failure rate and 30% to 40% less development time.
Innowise
Innowise says it has 3,500+ IT professionals, 60+ in-house medical software developers, 1,600 projects and 400 customers across 20+ countries. Its device work includes SaMD and software in medical devices for products intended for FDA and MDR markets. The company lists ISO 13485, ISO 9001 and ISO 27001, with services spanning embedded systems, wearables, patient and clinician apps, AI and cloud.
In a reported example, the team integrated Bluetooth Low Energy into a smart ECG monitor in 6 weeks, after a 5-day project start. Another project connected an AI wound-monitoring app to a Bluetooth thermal camera.
Softeq
Softeq has worked across hardware, embedded systems and software since 1997. It reports 200+ engineers and 22 locations, and lists ISO 13485:2016, ISO 27001:2022 and ISO 9001:2015. Its services include PCB and hardware design, embedded software, IoMT connectivity, mobile apps, cloud analytics, AI and remote patient monitoring.
A heart simulator upgrade reached a reported peak timing accuracy of 0.1 ms against a 0.5 ms admissible error across 11 modes. For Viit Health, a non-invasive glucose sensing prototype reduced the prototyping phase by 25%, according to the company.
Sigma Software
Sigma Software, which has provided IT services since 2002, reports ISO 13485:2016 certification in January 2025 and offices in 22 countries. Its medical and healthcare work includes device data and cloud analytics, remote monitoring analytics, embedded engineering and IoT. The company says its teams incorporate FDA, HIPAA and GDPR requirements into architecture and workflows.
A listed project migrated Siemens Healthineers’ CT scanner monitoring and usage analytics platform to Azure. Sigma Software reports that an 11-person team had been working on the project since April 2023.
SumatoSoft
Active since 2012, SumatoSoft reports 100+ specialists, 70% senior engineers and 350+ delivered solutions. Its services include IoMT software, wearable integrations, companion apps, Bluetooth connectivity, AI and cloud backends. It says its engineers work with ISO 13485, IEC 62304 and ISO 14971 processes, and lists ISO 27001 and ISO 9001.
Reported projects include a continuous glucose monitoring companion app that detected glucose events 60% faster than scheduled manual checks and shortened new-patient setup by 40%. A smart blood pressure cuff app had 20,000+ downloads and reported user satisfaction of 98%.
Questions to settle before kickoff
Can an existing prototype be used?
Often, but a working prototype may lack the requirements, risk records and traceability expected for a regulated product. A gap assessment can identify which code is suitable to keep, what needs refactoring and what should be rebuilt. Ask the team to document that assessment and connect retained components to the requirements and tests that support them.
Does software development need a separate regulatory consultant?
Not in every case. The team producing the software can often create much of the supporting material as the design and tests progress. Ask which documents the vendor will deliver, including requirements and design specifications, risk files and traceability matrices, and who will respond to reviewer questions. A specialist consultant can still help with pathway strategy and the final filing.
Who owns the code and project records?
The manufacturer should have the rights and access needed to maintain the device. The contract should address source code, design files, risk records, test evidence, repository access and build scripts, with documentation delivered in editable form. Without those materials, future maintenance may depend on the original supplier.
How should AI-assisted work be reviewed?
AI tools can help draft code, generate tests and update documentation, but engineers remain responsible for the output. Review, verification and traceability still apply. Ask how the vendor records its use of AI and how experienced engineers check generated work before it is included in the product.
A sound partner selection starts with the six stages and the records expected at each handoff. Compare vendors against the actual work your device needs, ask for evidence behind reported results, and settle ownership and review responsibilities before development begins. That gives your team a practical way to keep the software, its risks and its supporting records aligned from first scope through later updates.