Reveleer uses Google Gemini and Vertex AI to build a next-generation, AI-powered prospective risk adjustment solution that gives doctors the information they need to help patients at the point of care.
Reveleer uses Google Gemini and Vertex AI to build a next-generation, AI-powered prospective risk adjustment solution that gives doctors the information they need to help patients at the point of care.
Using AI to create healthcare intelligence that changes lives
In healthcare, no two patients are exactly alike. Therefore, providing the most valuable care requires a deep understanding of patients, their health concerns, and their goals. For more than a decade, Reveleer has helped healthcare organizations stay at the forefront of innovation and value-based care performance by using AI to gain a clear view of each patient by bringing together previously fragmented data.
“Doctors often have just a couple of minutes to review a patient’s history and understand the highest value care gaps before an appointment,” says Alan Tam, Chief Marketing Officer at Reveleer. “Giving them the most accurate, highest value, and easily digestible information at the point of care enables them to focus on the patient.”
One of the highlights of the Reveleer platform is its prospective risk adjustment solution. The hybrid AI-powered solution extracts and analyzes both structured and unstructured data from electronic health records, lab reports, doctor’s notes, and prescriptions from various data sources. It then delivers provable, explainable, audit-ready insights directly into a physician’s workflow, reducing provider abrasion and improving the patient experience and outcome.
For instance, a patient’s history might include some records that note hypertension symptoms, some with high blood pressure, and others that simply list changes in medications or other indicators of a chronic condition.
What we needed was a LLM that could really understand general English and deliver deeper insight into how past clinical records might indicate healthcare concerns. We found what we needed in Google Gemini.
Julien Brinas
SVP AI, Technology, Reveleer
One of the highlights of the Reveleer platform is its prospective risk adjustment solution. The hybrid AI-powered solution extracts and analyzes both structured and unstructured data from electronic health records, lab reports, doctor’s notes, and prescriptions from various data sources. It then delivers provable, explainable, audit-ready insights directly into a physician’s workflow, reducing provider abrasion and improving the patient experience and outcome.
For instance, a patient’s history might include some records that note hypertension symptoms, some with high blood pressure, and others that simply list changes in medications or other indicators of a chronic condition.
Reveleer recognizes and processes all of this information as evidence for a potential hypertension condition and summarizes it as a prediction indicator into a consistent patient history delivered into a physician’s workflow. This gives clinicians a better understanding of what issues a patient might have before meeting them. Clinicians can spend less time asking endless preliminary questions, and more time treating the patient.
“What we’re trying to do is predict health risks and back it up with accurate evidence from across a patient’s medical record. That’s the key to improving patient outcomes and reducing provider abrasion,” says Tam.
When Reveleer first started working with prospective risk adjustment suspecting, it was using natural language processing to read through unstructured data and search for health evidence, but the company figured it could provide even more insightful results and reduce false positives by building a next-generation engine using large language models (LLM). After testing many of the top models in the market, Reveleer chose Google Gemini thanks to its competitive balance of intelligence and cost to analyze up to tens of thousands of records per patient.
Finding the right model for the job with Gemini
Reveleer started its search for the right LLM with Vertex AI. Developers started experimenting with top first-party, third-party, and open source models through Model Garden on Vertex AI. At the end of these tests, developers determined that they needed the general understanding of the English language of Gemini, more than specialized healthcare tuning. Developers built a complex agent tasked with interpreting the clinical charts, using different versions of Gemini Flash and Gemini Flash-Lite for different actions throughout the agentic pipeline. By choosing the right model for each step, they built a robust pipeline capable of scaling to support growing patient medical records, while producing a step change in the level of accuracy interpreting both structured and unstructured clinical charts.
From what we’ve seen, Vertex AI is the most robust development platform for anyone wanting to work at scale with generative AI.
Julien Brinas
SVP AI, Technology, Reveleer
Reveleer makes extensive use of the Google Cloud ecosystem within its prospective risk adjustment solution for suspecting. Cloud Run is the foundation used to develop and run the AI. Data initially runs through Firestore for real-time analysis before it’s stored in BigQuery. Chosen for its flexibility and scalability, BigQuery currently stores all of the data needed to understand clinical records and how the AI interpreted and structured every piece of evidence used for suspecting every step of the way.
Throughout the development process, Reveleer worked closely with the Google Cloud team to bring the solution to life.
“Developing generative AI capabilities at scale is very difficult, especially when the output needs to be accurate and consistently reliable” says Julien Brinas, SVP AI, Technology at Reveleer. “Google Cloud is a great partner, collaborating with us to overcome engineering challenges and deliver fast, accurate performance.”
Turning hidden, unstructured data into high value clinical insights
Reveleer’s prospective suspecting capability consists of two core services. The first ingests all structured and unstructured data for a patient, including handwritten files. Patients can have thousands of chart files on average, each with half a dozen pages of content each. Reveleer’s evidence extraction agent carefully looks at all of the data across records, structured or unstructured, then translates the relevant pieces of evidence into a standard clinical language that guarantees that all evidences are described consistently.
Reveleer then analyzes these standardized clinical evidences to search for gaps in care and apply clinical reasoning using a deep set of clinical suspecting formulas developed by the clinical experts at Reveleer. For instance, pulling together medical history, labs, demographics, and symptoms from records coming from different clinics may bring to light new conditions as of yet unsuspected by the physician. Bringing together these disparate data sources into a single view better informs clinicians, allowing them to effectively determine whether additional testing might be appropriate.
While analysis of structured medical data is common, Gemini allows Reveleer to quickly and accurately pull data from valuable troves of unstructured data as well. As a result, Reveleer can deliver a much more comprehensive picture of the patient’s conditions and therefore much more value to help healthcare organizations improve patient care.
All of this work has enabled the following use cases:
- Provable and explainable results: Suspect recommendations are grounded in transparent, deterministic clinical logic, ensuring every suspect is traceable, audit-ready, and fully supported by direct-source documentation on the new Reason field. This builds trust and confidence for providers and compliance teams, enabling at a glance decision-making.
- Reduced provider abrasion: By delivering fewer, more relevant suspects—with up to 50% less noise than legacy NLP—provider and coding teams spend less time navigating irrelevant findings and more time focused on care, not compliance.
- Rapid adaptability and customization: Reveleer’s deterministic formulas and agentic AI enable customers to update logic, filters, and models in minutes instead of months, ensuring fast alignment with regulatory, clinical, or local needs.
- Agentic evidence extraction: Complex generative AI translates structured and unstructured medical charts into a standardized clinical corpus.
- Continuous improvement: The engine is enhanced by clinician feedback and self-correcting AI agents—formulas and evidence rules evolve rapidly, reflecting best practices and regulatory changes.
With these innovations, Reveleer is setting a new standard for accuracy, transparency, and efficiency in prospective risk suspecting and value-based care. And this applies to everyone up and down the chain—from practitioners having better at-a-glance information for patient care decisions to flexible workflows for coding and compliance teams, Reveleer is driving improvements across the board.
Reveleer also has the potential to build upon the standardized clinical evidence data with new use cases. “It’s almost like we have a magic wand that lets us understand unstructured clinical data. It’s going to power a lot of Reveleer services moving forward,” says Brinas.
Gemini Flash and Flash-Lite unlock so many new use cases and capabilities for us. We’re beyond excited at what this means for Reveleer and our customers.
Alan Tam
Chief Marketing Officer, Reveleer
As a pioneer in value-based care enablement, Reveleer is purpose-built to solve the most pressing real-world challenges faced by provider and health-plan organizations today.
Industry: Healthcare and Life Sciences
Location: United States
Products: Vertex AI, Gemini, Cloud Run, BigQuery, Firestore, Cloud Composer
