Industries · Healthcare

Healthcare software from a team that runs its own clinical product.

CS2 builds and operates Notermed, an AI medical scribe developed with a practicing US physician. We have already worked through HIPAA agreements, patient data boundaries and the moment a doctor signs a note. That same healthcare software development goes to clinics, health-tech startups and healthcare service providers in Canada and the US.

Live products
4
Live REST endpoints
130+
Incorporated
2022
When you need this

Problems we hear from clinics and health-tech teams

Healthcare software fails on the details nobody budgeted for: where the audio goes, who signed the agreement, what the AI is allowed to write. These are the situations we get called about.

You want AI in a clinical workflow and the compliance questions arrived first

Where does the recording go, which vendor has signed a business associate agreement, what must the model never write into a note. Those decisions have to be made before the first line of code. We made them for our own product.

Your clinic runs on an EHR and a dozen tools that do not talk to each other

Scheduling, billing, fax, the patient portal, the lab interface. Staff will not adopt another login. New software has to fit what they already use.

A health-tech startup has to ship without a compliance incident

You have a clinical advisor, a deadline and a pilot clinic waiting. You need engineers who have shipped patient-facing software and know which shortcuts are safe and which ones end the pilot.

Patient data has to stay in Canada

Ontario PHIPA, a provincial hosting rule or a hospital procurement checklist says Canadian residency. The architecture has to answer that on day one, including the AI providers in the chain.

A change at your AI vendor must not stop the clinic

Pricing, policy or the vendor agreement can change at any AI provider. If your product is wired directly to one vendor SDK, that change becomes a rebuild. Notermed went through this move, and we wrote it up.

You need an audit trail you can actually show

Who opened a record, who edited the AI draft, when the note was signed. Regulators and hospital partners ask. The system has to answer without a developer digging through logs.

What we build

Healthcare software, from the scribe to the hosting

We build for clinics and clinic groups, for health-tech startups, and for service providers such as billing, transcription and home-care companies.

AI medical documentation and scribing

Speech-to-text and note generation with instructions that forbid invented findings, a clinician review-and-sign step, and a fallback to standard behaviour when a custom setting breaks. This is the pattern running in Notermed today.

Patient and clinician portals

Auth-gated web and mobile applications with role-based access, consent capture and per-tenant data separation that holds inside the AI features too, not only in the database.

Integration with existing practice tooling

Connections to your EHR and practice-management system through the vendor API, HL7 or FHIR interfaces where offered, or structured file exchange where that is all there is. We fit what you have before proposing to replace anything.

HIPAA, PHIPA and PIPEDA as engineering practice

Business associate agreements with every vendor that touches patient data, encryption at rest and in transit, least-privilege access, retention rules, logged access. This is what HIPAA compliant software means in practice: we describe what we do and do not sell certifications we do not hold.

Provider-agnostic AI

The model sits behind an interface your code owns. Notermed moved from OpenAI to Amazon Bedrock as a short project rather than a rewrite, and the old provider can no longer even be configured. Client systems get the same design.

Hosting and data residency in Canada

AWS Canadian regions, infrastructure as code, audit logging and backups you can show an auditor. US deployments where your patients and agreements are in the US, which is how Notermed itself runs.

Proof

A clinical product we operate, and the work behind it

We are not describing a hypothetical. Notermed is a CS2 product, wholly owned, with physicians using it to write their notes.

Product, wholly owned

Notermed, AI medical scribe

Built with a practicing US physician. The doctor records the visit, the app drafts the note, and the doctor reviews, edits and signs it. Nothing enters the chart before the signature. Patient data stays with HIPAA-covered providers under a signed agreement and is never used to train models. Clinics can start free.

Visit notermed.com

Case study

Notermed: marketing site and app on one codebase

Java 17 Spring Boot API with Spring Security and JWT, PostgreSQL with pgvector, Redis, notes streamed to the clinician over WebSockets, AI through AWS Bedrock via Spring AI, a dedicated Whisper speech-to-text service and a React Native mobile app.

Read the case study

Engineering note

Switching AI providers in a medical app

Seven lessons from moving Notermed onto HIPAA-covered AI: making the old provider impossible to configure, treating a model change like a data migration, testing with real documents, and writing down what the AI must never do.

Read the article

Past performance

GDPR-regulated systems for Vodafone Kabel Deutschland

Three years (10/2022 to 10/2025) as a CS2 engagement on the field-operations REST backend of a German Tier-1 telecom, processing personal data of field technicians and subcontractors under the GDPR. Java, Spring Boot, Kubernetes, AWS.

See past performance
How we work

From data map to operations

  1. Call and data map

    Thirty minutes on what you want to build, then a map of where patient data flows today: systems, vendors, agreements, residency.

  2. Scope in writing

    A written scope with milestones and a price. Fixed scope for a defined build, a dedicated team for ongoing product work. Nothing is invented on a call.

  3. Build in short releases

    Working software every few weeks, reviewed by a clinician on your side. Security and audit logging are in the first release, not the last.

  4. Operate

    Monitoring, backups, audit logs and vendor changes handled by the people who built it, or a documented hand-off to your team.

Stack

What our healthcare work runs on

  • Java 17 + Spring Boot
  • Spring AI + Amazon Bedrock
  • Whisper speech-to-text
  • Next.js + TypeScript
  • React Native (Expo)
  • PostgreSQL + pgvector
  • Redis + WebSockets
  • Spring Security, OAuth 2.1, JWT
  • AWS (Canada and US regions)
  • S3 + CloudFront
  • Terraform
FAQ

Questions clinics and founders ask us

Are you HIPAA certified?

Nobody is. HIPAA has no official certification, and we will not claim one. What we can describe is our practice: business associate agreements with every vendor that touches patient data, encryption at rest and in transit, access controls, audit logs, and the same controls in our own product. The same goes for Ontario PHIPA and PIPEDA: we build to them and say so plainly, without a certificate we do not hold.

Can patient data stay in Canada?

Yes. We deploy to AWS Canadian regions and keep the AI providers inside the same boundary. For US clients, data stays in the US. Notermed runs that way, with HIPAA-covered providers under a signed agreement.

Which EHR systems do you integrate with?

It depends on what the vendor exposes. We work with the EHR vendor API, with HL7 or FHIR interfaces where they are offered, or with structured export and import when that is all there is. We confirm this with your vendor during the first call, before anything is scoped.

What happens if our AI vendor changes its terms?

The model sits behind an interface your code owns, so a vendor change is a configuration and testing project rather than a rewrite. We did exactly this when Notermed moved from OpenAI to Amazon Bedrock, and the article on this site walks through what it took.

Do you work with startups as well as clinics?

Yes. Clinics and clinic groups, health-tech startups building a product, and service providers such as billing, transcription and home-care companies that need software built or extended.

How do you price healthcare projects?

You get a written scope with a price after the first call. Defined builds are fixed scope; ongoing product work runs as a dedicated team billed monthly. We do not publish prices because no two clinical setups are alike, and we will not invent a number to win a call.

Who does the work?

Senior-led delivery, scaled to the project. CS2’s senior engineers lead the work with vetted developers, and one account owner handles the relationship from the first call on. The people on your project are the people who run Notermed.

Tell us about the clinic or the product.

Thirty minutes is enough to map where patient data flows today and what you want to build. You will get a scope you can act on afterwards, not a sales deck.