Goodside Health, the management company for a series of PT practices, is operating across multiple clinic locations, with both in-person and remote patient care. As the business grew, so did the complexity of its data. Financial records sat across multiple instances of QuickBooks and Ramp. Clinical records lived in Prompt EMR, an EMR system purpose-built for physical therapy. Their CRM was an in-house Notion build. Referrals came in by fax.
Each system worked in isolation. None of them talked to each other.
The goal Goodside Health came to us with was straightforward in principle: a single source of truth that could support financial reporting, day-to-day clinic operations, and eventually automate the manual processes their tooling left unhandled.
The situation
Goodside Health's reporting was being held together by one of the partners, an ex-private equity operator with a serious command of Excel. The reporting worked. It surfaced the metrics that mattered. But it consumed a significant portion of his time every week, and the cracks were starting to show.
Excel has hard row limits. As the business scaled, they were bumping against them. More importantly, when the same metric lives in multiple spreadsheets maintained by different people, inconsistencies creep in. One sheet uses business days, another uses calendar days. A number gets updated in one place but not another. Over time, leadership loses confidence in the data, and the reporting that was supposed to drive decisions becomes something people argue about instead.
The two managing partners had a clear division of focus. One owned the business and financial side. The other came from a clinical background and was primarily focused on clinical process, care quality and patient outcomes. The reporting burden was landing entirely on one of them, pulling him away from higher-level work. The other had no clean visibility into operational data relevant to clinic management.
The challenges
HIPAA compliance, disconnected data, vendor dependencies, and governance at scale were not sequential problems to solve. They all had to be addressed together, in a system that was still growing while we were building it.
HIPAA compliance without the enterprise tax
US healthcare data regulation sets out a clear baseline: encryption at rest, access logging, audit trails, documented data handling. Most of that is standard engineering practice in 2026. The problem is that vendors know healthcare companies have no choice but to comply, and many price accordingly. We looked at several tools where the only thing the enterprise tier was selling was a BAA signature and a logo on a compliance page. Some wanted north of $10,000 a month for that. The underlying platform was identical.
The build versus buy calculation was central to the architecture decisions we made here.
Disconnected data and complex business logic
Connecting these systems is not a simple data job. Healthcare operations generate hundreds of metrics, and many of them require understanding how the business actually runs before you can model them correctly. Outpatient PT has its own domain logic. Add remote treatment, multiple clinic locations, and metrics that need to roll up differently depending on whether you're a clinic manager or a network-level partner, and you have a significant modeling problem.
Our analyst has been embedded in this data for six months and is still uncovering nuance. That is not a criticism of the work. It is a realistic picture of what complex healthcare data looks like when you get close to it.
Governance at scale
As the business grows and acquires more practices, data governance becomes load-bearing. Access controls need to reflect org structure. Deletions need to cascade correctly and be logged, which matters both for HIPAA and for CCPA compliance for California patients. Metrics need to be versioned so that when a definition changes, you know when it changed and why. Building this in from the start, while the business is still growing quickly, was harder than building it into a stable system. But retrofitting governance into a mature data platform is significantly worse.
Vendor data quality
A meaningful portion of Goodside Health's data comes from third-party systems they do not control. Vendors change APIs, update schemas, and modify data formats, sometimes without notice. That is a real operational risk for any data platform that depends on those feeds. It required building resilience into the pipeline rather than assuming clean, stable inputs.
The solution
We aligned on Azure as the cloud foundation. Microsoft's compliance posture is well-suited to regulated healthcare environments, and the tooling around data governance on Azure is mature. Databricks sits at the core of the platform, handling both the data lake and the compute layer. The choice of Databricks was partly about scalability and partly about giving the partners, both of whom have strong analytical backgrounds, a platform that could support ad hoc exploration on top of standard reporting.
Dagster handles orchestration. It coordinates the pipelines that pull from each source system and ensures data flows in the right order with proper dependency management.
For the modeling layer, we used dbt. The practical benefit here is twofold: it generates automated tests against the data, and it creates a version-controlled record of every metric definition. When a metric changes, you can see exactly what changed and when. That matters a lot in an environment where leadership debates report numbers.
n8n handles workflow automation and was self-hosted. The automation layer is where the longer-term value starts to compound, covering referral follow-up, scheduling reminders, and the gaps between systems that the EMR does not fill natively. That work is ongoing and will be the subject of a separate case study.
For the compliance requirements that genuinely required vendor tooling, we used Azure-native features where possible rather than paying enterprise premiums to third-party tools offering little beyond a signed agreement.
The results
The reporting that previously consumed a significant portion of a partner's week now runs automatically and distributes to the right people at the right cadence. That time has been redirected toward business development and clinic-level operational work.
The data foundation also unblocked Goodside Health's expansion into remote patient care and telehealth. Connecting the clinical and operational data across modalities required a layer of infrastructure that simply did not exist before. With the data lake in place, those connections became possible.
Governance built into the platform means that when a metric is wrong, it gets caught. Misaligned definitions, vendor-side data changes, and edge cases in business logic now surface as alerts rather than as errors in a board report.
The referral automation work, now in progress, was only possible because the data foundation was in place first. Automating a process you do not fully understand yet is how you automate the wrong thing. Getting the data right first was the correct sequence.
“We came to Jinka with a serious data infrastructure challenge: a HIPAA-compliant data lake, real-time reporting, and API-driven automations all at once. They delivered on all of it. Reports that used to take hours now run instantly, automations have reduced the administrative burden on our team, and we finally have clear visibility into how decisions affect the organization downstream. On top of that, they helped us build out the front and backend of a separate software product entirely. That range says a lot about who they are. Easy recommendation.”
Building a data foundation for a regulated business?
We design and build production-grade data platforms for healthcare and other complex domains.