Introducing Meridic

For every hour a physician spends with a patient, they spend two hours doing paperwork. Prior authorization requests. SOAP notes. Referral letters. Insurance appeals. This ratio has not improved in a decade — it has gotten worse. Physicians are not burning out because medicine is hard. They are burning out because the administrative systems surrounding medicine are failing them, and the tools sold as solutions have not changed the equation.

The people who chose medicine — who spent a decade training to diagnose, treat, and care — are now spending the majority of their working hours navigating insurance portals and dictating notes into systems designed in the 1990s. This is not a clinical problem. It is a systems problem. And systems problems have engineering solutions.

Why the First Wave of Healthcare AI Has Failed Clinicians

The technology industry's response to the administrative crisis in medicine has been to ship chatbots. Tools that can answer a clinical question, or generate a note, or suggest a differential diagnosis — but not all three, and certainly not in the context of a real patient workflow. Single-model tools designed for single tasks. They integrate with nothing. They know nothing about the patient in front of you. And they were not built for the stakes of clinical decision-making.

The Problem Is Not the Model. It Is the Architecture.

“No single AI model is best at everything clinical work demands. The breakthrough is not a better model — it is an orchestration layer that deploys the right model for each task, with the right clinical context, every time.”

Clinical reasoning — working through a differential, evaluating evidence, weighting risk — requires a model trained specifically on medical literature, with validated safety benchmarks. Long-form documentation — prior auth appeal letters, specialist referrals, discharge summaries — requires a model with exceptional language fluency and the ability to hold complex context across thousands of tokens. Structured data extraction — parsing an ICD-10 from a denial letter, normalizing lab results from an EHR export, pulling dates from a PDF attachment — requires a different capability entirely. These are not the same problem. Treating them as the same problem is why every clinical AI tool built so far has underperformed.

Introducing Meridic: AI Orchestration for Clinical Teams

Meridic is a multi-model AI orchestration platform purpose-built for clinical workflows. It is not a model. It is the conductor. Meridic routes each task to the model best equipped to handle it: Google MedGemma for diagnostic analysis and clinical reasoning, Anthropic Claude for long-form documentation and prior authorization letters, Google Gemini for structured data extraction from EHR records and attachments, and OpenAI GPT for general informational queries. Each model does what it was built to do. The orchestration layer — not any individual model — is where the clinical value is created.

One Platform. Every Clinical Context.

Every other AI tool in healthcare starts and ends at the EHR. But clinical work does not live only in the EHR. A prior authorization denial arrives in your Gmail. The specialist's letter is a PDF in your Google Drive folder. The patient's insurance information is in an Outlook calendar invite from your biller. Referral PDFs land in inboxes. Pharmacy rejections come through email. Meridic connects all of it. Personal connectors — Gmail, Google Drive, and Microsoft Outlook via OAuth — give Meridic context from the communications and documents that surround clinical care. Clinical connectors — Epic, Cerner, and Athenahealth via SMART on FHIR — give Meridic the structured clinical record. No other tool in this space integrates both. Meridic does, because the workflows that actually consume physician time live at the intersection of both.

A Morning, Reimagined

Consider what Meridic does before your first patient walks in. Overnight, a prior authorization denial arrived in your Gmail. Meridic read it, identified the denial code, pulled the relevant medication history and diagnosis records from Epic, and drafted an appeal letter using Claude — with MedGemma's diagnostic reasoning embedded as supporting clinical justification. The letter is ready for your review. One click submits it to the payer portal and sends you a confirmation. Your morning documentation templates are pre-populated with each patient's relevant history from the last visit. The lab result that arrived at 11pm — flagged as outside reference range — has been added to your task queue with the relevant clinical context and a suggested follow-up action. You walk in at 8am. The administrative work that used to define your morning is done.

import { Meridic } from "@meridic/sdk";

const client = new Meridic({ apiKey: process.env.MERIDIC_API_KEY });

// Trigger a prior authorization appeal workflow
const result = await client.workflows.run({
  type: "prior-authorization",
  patientId: "pt_92f14a",
  connectors: {
    ehr: "epic",          // Pull diagnosis codes, medication history, recent labs
    email: "gmail",       // Read denial letter, send submission confirmation
  },
  routing: {
    reasoning:  "medgemma-27b",     // Clinical justification & diagnostic context
    drafting:   "claude-opus-4-6",  // Long-form appeal letter generation
    extraction: "gemini-2.5-pro",   // Structured data from EHR and PDF attachments
  },
  context: {
    denialCode: "CO-97",
    icd10: "Z79.4",       // Long-term current use of insulin
    notes: "Appeal denial. Reference A1C trend (18 months) and endocrinologist letter.",
  },
});

console.log(result.status);    // "submitted"
console.log(result.draftUrl);  // Preview and edit the letter before sending
console.log(result.timeMs);    // 4312 — completed in 4.3 seconds

Built for Clinical Teams, Not Consumers

Meridic is designed for deployment across clinical organizations — practices, health systems, and specialty groups. It is HIPAA-compliant by default, with Business Associate Agreements available across all three major cloud providers: Google Cloud Platform, Microsoft Azure, and Amazon Web Services. No patient data leaves your compliance boundary. Multi-cloud deployment eliminates single points of failure and prevents vendor lock-in. Your data governance requirements are met before the first workflow runs. The product is built for teams, not individuals — with role-based access, audit logs, and administrative dashboards designed for clinical operations.

“Healthcare's administrative burden is a solvable engineering problem. We built Meridic to solve it.”

Join Early Access

We are currently onboarding a limited cohort of clinical teams — primary care practices, specialty groups, and health systems ready to operate differently. If your team is spending more time on administration than on patients, Meridic was built for you. Apply for early access today. We will respond within 48 hours.