Service · S/01 Healthcare Infrastructure

HIPAA cloud architecture, built around the PHI boundary.

We architect HIPAA-aligned cloud environments for hospital systems, clinical SaaS, pharmacy benefit managers, biotech, and digital health platforms. PHI boundary design, encryption posture, identity federation, audit logging, and Terraform IaC across AWS, GCP, Azure, and OCI.

EngagementFixed-fee build
or audit + roadmap
Timeline2 weeks (audit)
8 weeks (build)
FrameworksHIPAA · HITRUST
SOC 2 (overlapping)
CloudsAWS · GCP · Azure · OCI
w/ BAA in place
01 — The Problem

Healthcare engineering teams pay for the cloud twice. Once to build, once to prove.

The HIPAA Security Rule does not specify cloud services. It specifies safeguards. Teams who interpret it as a checklist of services end up paying for the architecture twice: once when they build it the easy way, and again when they rebuild it to satisfy a HITRUST assessor or a hospital procurement reviewer.

We have built and audited HIPAA cloud environments for clinical SaaS startups selling into hospital systems, pharmacy benefit managers running production PHI workloads, and digital health platforms preparing for HITRUST. The pattern is the same in every engagement: the engineering team understands the application, the compliance team knows what HIPAA wants, and the cloud architecture in the middle has never been built to satisfy both.

Stonebridge designs the environment around the PHI boundary first. Encryption, identity, logging, and network architecture are derived from where PHI flows, not bolted on after.

02 — Control Mapping

HIPAA Security Rule safeguards the environment must satisfy.

The HIPAA Security Rule (45 CFR § 164.308 through § 164.312) names safeguards, not services. A correctly-architected HIPAA cloud satisfies the technical safeguards through how it is built, not through policies that describe what it is supposed to do.

The safeguards the environment most directly touches:

  • §164.308(a)(3)Workforce security. Identity federation with least-privilege IAM and time-bounded access to PHI-bearing systems.
  • §164.308(a)(4)Information access management. Role-based access boundaries, with PHI access audited and reviewable.
  • §164.312(a)(1)Access control. Unique user identification, automatic logoff, and encryption of data at rest.
  • §164.312(b)Audit controls. Centralized, tamper-evident logging of access to PHI and changes to systems that touch PHI.
  • §164.312(c)(1)Integrity. Signed artifacts, immutable infrastructure, and drift detection on production state.
  • §164.312(e)(1)Transmission security. Encrypted in transit with mutually-authenticated, modern TLS for all PHI flows.
03 — The Architecture Pattern

Build around the PHI boundary, not the application boundary.

The architecture pattern that survives HIPAA scrutiny rests on four principles. Implementation varies across AWS, GCP, Azure, and OCI; the architecture is the same.

Principle 1: Identify the PHI boundary first

Before any Terraform is written, we map where PHI enters the system, where it is stored, and where it leaves. Every account, network range, and service that touches PHI is inside the boundary. Everything else is outside. This single decision drives every subsequent architectural choice.

Principle 2: Account-level isolation, not tag-level

PHI-bearing workloads live in separate cloud accounts (AWS), projects (GCP), or subscriptions (Azure) from non-PHI workloads. Identity federates from a central IdP with role-based, time-bounded access. Network connectivity between accounts is explicit, mutually-authenticated, and audited.

Principle 3: Customer-managed keys for PHI-bearing storage

KMS-backed encryption with customer-managed keys for every PHI-bearing data store. Key usage is logged. Key rotation is automated. The blast radius of a single key compromise is bounded by design.

Principle 4: Centralized, retention-locked audit logging

Every account, service, and identity event ships to a centralized logging account that engineers cannot modify. Retention locks (S3 Object Lock or equivalent) prevent tampering. Logging is not a feature; it is the system of record auditors will use to validate every other control.

04 — What You Get

Six artifacts your auditor will ask for.

A HIPAA Cloud Build engagement produces a defined set of deliverables, each mapped to safeguards and ready for assessment package inclusion.

D/01

PHI boundary architecture

  • PHI boundary diagram
  • Data flow diagrams
  • Network topology
  • BAA-covered service inventory
D/02

Infrastructure as code

  • Terraform modules for the boundary
  • OPA / Sentinel policy gates
  • Drift detection wired to alerting
  • State backend in the PHI boundary
D/03

Identity & access

  • IdP federation
  • Role catalog with least-privilege scopes
  • Access review automation
  • Break-glass procedure
D/04

Encryption & key management

  • KMS topology
  • Customer-managed keys for PHI
  • Automated rotation
  • Key usage audit logging
D/05

Logging & monitoring

  • Centralized logging account
  • Retention locks
  • SIEM integration
  • HIPAA-mapped dashboards
D/06

Control documentation

  • HIPAA safeguard narratives
  • Continuous monitoring plan
  • Incident response runbooks
  • HITRUST CSF crosswalk
05 — Common Mistakes

Five patterns that fail HIPAA assessments.

We see the same mistakes repeatedly when healthcare engineering teams build HIPAA cloud environments without architectural help.

  1. One account, all PHI

    A single AWS account or GCP project hosting production PHI, dev environments, and corporate workloads. The first IAM mistake exposes everything. Separate accounts for PHI workloads; explicit, audited cross-account access.

  2. Default encryption, no key rotation

    Relying on default cloud-managed keys with no rotation discipline. The encryption is in place but the key management discipline is not. Customer-managed keys with automated rotation, logged usage, and clear key ownership.

  3. BAA not enforced in IaC

    A BAA covers specific services; the IaC happily provisions services that are not BAA-eligible. Policy gates that reject non-eligible services at plan time prevent the silent boundary breach.

  4. Logs that engineers can edit

    Audit logs in the same account engineers operate in. Logging needs its own account, retention locks, and access boundary distinct from production engineering.

  5. HIPAA as a policy document, not an architecture

    Policies describing what the environment should do, with infrastructure that does not actually do it. We write control narratives alongside the Terraform; the policy is generated from the architecture, not the other way around.

06 — Engagement

Two ways to engage. Fixed scope, fixed price.

Most clients start with the audit. It produces a written roadmap that often gets approved as a follow-on build. Clients with a fixed audit deadline come straight to the build.

07 — Recent Work

A representative engagement.

One of our healthcare engagements involved auditing a clinical platform's GCP infrastructure to identify HIPAA gaps before a third-party assessment, then codifying the baseline so future deployments could not regress.

HEALTHCARE · GCP

HIPAA GCP infrastructure audit, ahead of third-party assessment.

The platform was running production workloads on GCP with PHI in flight. The compliance team had a required version baseline for Tomcat, JavaScript runtimes, and operating systems. Manually inventorying every VM against the baseline would have taken weeks.

We wrote an Ansible and Python tool that connected to each VM through GCP Identity-Aware Proxy, scraped versions, and produced a compliance matrix mapping current state to the required baseline. From there we extended the work into the pipeline, codifying the baseline as a policy gate so future deployments could not regress.

~85%Reduction in inventory time
vs manual process
PassedFollow-up audit cleared
on first-party review
ContinuousBaseline enforcement
codified in the pipeline
08 — Questions

Frequently asked, directly answered.

Q/01Do you sign Business Associate Agreements?
Yes. Stonebridge signs BAAs as part of every healthcare engagement where PHI may be reached during the work. We treat the BAA as a precondition for access, not a paperwork step at the end.
Q/02Which clouds do you support for HIPAA workloads?
AWS, Google Cloud, and Azure all have HIPAA-eligible services and BAA paths. Oracle Cloud Infrastructure also supports HIPAA workloads. We architect on whichever fits your customer requirements, BAA constraints, and existing footprint. We will not recommend a cloud migration unless your current cloud genuinely cannot meet the requirements.
Q/03Does HIPAA require us to use a specific cloud service?
No. HIPAA names safeguards, not services. What matters is whether the services you use are covered by your cloud provider's BAA and whether they are configured to satisfy the technical safeguards. We help you choose the smallest set of HIPAA-eligible services that meets your workload requirements.
Q/04What about HITRUST?
HITRUST CSF is the prescriptive framework most healthcare buyers will eventually ask for. The cloud architecture pattern that satisfies HIPAA covers most HITRUST CSF technical controls; what differs is the mapping rigor and the third-party assessment cadence. We deliver a control mapping document with each engagement that maps environment outputs to HITRUST CSF where relevant.
Q/05Can you work with our existing BAA-covered infrastructure?
Yes. Most engagements modernize an existing environment rather than rebuild from scratch. We assume your BAA scope, your existing accounts, and your existing tooling are the starting point and we evolve them. Greenfield is the exception.
Q/06How quickly can we be audit-ready?
A typical HIPAA Cloud Build is eight weeks: two weeks of discovery, five weeks of build, one week of handoff and runbooks. If you have a fixed audit or BAA deadline, we plan backwards from that date and align the milestones accordingly.

Ship to production. Pass the HIPAA assessment.

Most discovery calls take 30 minutes. We come back with a written proposal within 48 hours. If we are not the right fit for the engagement, we will tell you in the first call and point you somewhere that is.

Book a 30-minute call