Medical Device
Software as a Medical Device, connected devices, and companion apps, built inside design controls.
Services
Three ways we help health tech teams ship.
Product strategy and product development grounded in regulated health products, and a lean, AI-assisted way of building that we proved on our own work first.
Service 01
Product Management & Strategy
We help health tech teams decide what to build and why, then get marketing, regulatory, quality, clinical, and engineering teams behind the plan.
Product strategy and roadmaps
A strategy and roadmap leadership can fund and engineering can deliver.
User and market research
Interviews with patients, clinicians, and buyers, plus competitive and market landscape work.
Business cases and positioning
The case for a new product or platform, written for the people who approve the budget.
Requirements and product definition
Product requirements, user stories, and acceptance criteria that hold up in a regulated environment.
Service 02
Product Development & Engineering Management
We lead the build, from the first requirement to launch, inside the constraints of medical device, EMR, and clinical research environments.
New product delivery, 0 to 1
Concept through design, build, and launch, with the cross-functional alignment a regulated launch needs.
Engineering management
Leading or rebuilding engineering teams, including global and vendor teams, with a clear process and a predictable delivery cadence.
Vendor selection
Choosing and onboarding engineering partners, from the buyer's side of the table.
Regulated delivery
Traceability and documentation that hold up under FDA, EU MDR, ISO 13485, HIPAA, GDPR, and GxP.
Domain expertise
Where we have shipped before.
Digital Health
Patient and clinician apps, digital therapeutics, and app-prescribing programs that reach real users at scale.
Electronic Medical Records
Apps that live inside the clinical workflow, including SMART on FHIR launches in Epic.
Real-World Data
Real-world data platforms, NLP pipelines, and evidence generation from clinical and claims data.
Clinical Trials
Trial technology from EDC integration to patient recruitment and eligibility matching.
Service 03
Lean AI Delivery
An AI-assisted software lifecycle, set up so every model call earns its cost.
The problem
Most teams adopted AI tools one person at a time. On our own build we measured where the tokens went: over 99% were agents re-reading context, and under 1% were spent writing anything. Cost grows with how many turns an agent takes and how much it re-reads each turn, not with how much it produces. Most teams can't see that, so they can't fix it.
Clear lanes for AI agents
Product, design, engineering, and project management each get their own agent and scope, so one lane's output is the next lane's input instead of a duplicate. Each lane owns its own files and its own queue. Work that crosses a lane goes through a ticket, not a side conversation.
One shared project context
A versioned coordination file that every agent and person reads before working, with ownership rules so nobody overwrites another lane. Kept short on purpose, because every agent re-reads it on every turn.
One session per task
Agents self-restart fresh for each task instead of running all week, which keeps the context they re-read small.
Requirements before code
Specifications and acceptance criteria are agreed before anything is built, so agents don't build the wrong thing and then rebuild it.
Decision records
Architecture and design decisions are logged once, so agents stop reopening them.
Evidence gates
Human-in-the-loop review gates at each stage, with the traceability regulated teams need.
Rules enforced, not just written
Where a written rule didn't hold, we enforce it in the tooling: guards that stop agents from sweeping each other's work into a commit, discarding shared changes, or starting runaway sessions.
Cost per point delivered
A baseline of where your AI spend goes today and the right model for each kind of task. Every sprint has a baseline of tokens used against story points shipped, so you can see whether a process change made delivery cheaper.
Assess
Map the AI tools in use across your lifecycle, who uses them, and what they cost.
Design
Define agent lanes, the shared context, decision records, and review gates for your team.
Pilot
Run it on one product team and measure output against spend.
Hand over
Document the playbook and train your leads to run and extend it.
Proven on our own work. StudyMatcher is built this way: six AI agent lanes, one human approver and weekly sprints, all tracked in GitHub issues, with token use measured against points delivered every sprint.
How we engage
Pick the shape that fits the problem.
Fractional leader
Part-time product or engineering leadership embedded in your team for an extended period.
Defined engagement
A scoped project with a clear deliverable, such as a product strategy, a requirements set, or a vendor selection.
Advisory
A standing working session with your leadership team to pressure-test plans and decisions.
Start a conversation
Tell us what you are building.
Where the product is today, where it needs to go, and what is in the way. Michael replies personally.