The Albert Einstein Hospital is Latin America's top-ranked hospital, serving both public and private patients with a commitment to equal quality of care. But in radiology, the gap between having data and acting on it was costing physicians time they could not afford to lose. Clinicians were reviewing free-text reports from multiple systems, one case at a time, and manually building patient cohorts just to identify who needed follow-up. With Genie on the Databricks Data + AI Platform, the hospital gave physicians governed, natural-language access to clinical data, replacing weeks of manual work with answers in seconds and sparking adoption across specialties.
When physicians had data but no practical way to access it
At Albert Einstein Hospital, delivering equal care to public and private patients depends on turning clinical information into action, and in radiology, that still requires significant manual effort. The hospital processes more than 120,000 imaging exams a year, generating free-text reports across different systems and formats.
To build patient cohorts and identify people who needed follow-up consultations, clinicians reviewed those reports one case at a time, pulling results into a spreadsheet to assemble the clinical picture needed for decision-making. With roughly 3,000 patients per month moving through the diagnostic workflow and 10,000 physicians across the hospital, a single-cohort analysis could take a week or longer.
The data existed on AWS, but accessing it meant stepping out of the clinical workflow, navigating the AWS Console and writing SQL to pull it. The interface was complex, the cost per query was high and most physicians were not proficient in the system.
"Some of our physicians joked that they would need to go back to school just to query their own data," said Thiago De Souza, Data Platform Manager at Hospital Albert Einstein.
Brazilian healthcare data falls under the Lei Geral de Proteção de Dados Pessoais (LGPD), the country's data protection law comparable to the GDPR, and Einstein's older environment made it difficult to cleanly limit access by role. Physicians, operators and coordinators needed very different views of the same information, but the tooling could not enforce those boundaries at scale.
"The governance mindset was already here. What we lacked was a platform that could enforce it," Thiago said.
Genie gives physicians secure, self-service access to radiology insights
Einstein started by putting Genie at the front door of radiology data on the Databricks Platform, beginning with a 10-physician pilot built around real clinical questions so the team could refine prompts with live users before expanding. That early prompt work turned out to be the main lift, because once it was in place, the experience felt intuitive even for non-technical clinicians.
"Physicians don't write queries," Thiago said. "They ask a question the way they would ask a colleague, and Genie brings back the answer."
The governance layer behind Genie was Unity Catalog, which Einstein connected to its existing Active Directory, so role-based permissions carried over without having to rebuild security from scratch.
The hospital created three Genie spaces for physicians, operators and coordinators, each with permissions matched to the role. Operators are blocked from personal health data, physicians cannot see certain payment details and coordinators can access broader information when needed. The governance model Einstein had always wanted was now enforceable out of the box.
"Unity Catalog took our governance to a level we couldn't reach before,” said Thiago. “Once you operate that way, you can't imagine going back."
With that foundation in place, Genie quickly moved beyond basic lookups. Physicians now request focused reports in response to a single prompt. They can also build advanced analyses that intersect datasets, match records across exam types and calculate intervals between procedures to understand care patterns and physician workload.
Faster answers and time back to spend with patients
Questions that once took a week or even a month of manual spreadsheet work now return answers in seconds, and onboarding new clinicians to the workflow has dropped from one month to one week. Einstein grew Genie from 10 pilot physicians to 110 active users, with adoption spreading evenly among younger and older doctors, including physicians in their seventies.
Beyond the speed gains, analysts can focus on higher-value work rather than assembling records by hand, and physicians can redirect expensive clinical time toward patient care and hospital operations. In a setting where physician time carries both clinical and financial weight, that reclaimed capacity has clear operational value.
"Every minute a physician spends pulling data is a minute taken from a patient," Thiago said. "Genie gave us those minutes back."
Einstein is already using that momentum to widen the impact. A physician council is exploring how data can improve radiology, pathology and laboratory processes. At the same time, the hospital looks ahead to connecting radiology, pathology and genetic data for oncology and rare-condition identification.
"We started by giving physicians faster answers, but what we're building toward is something bigger,” said Thiago. “Discovering patterns no one has looked for yet. That's the next chapter for Einstein, and Databricks is how we get there."