The challenge
Patients with common non-emergency conditions face four bad choices: a primary care physician booked weeks out, urgent care that's crowded and expensive, an ER trip for the wrong reasons, or skipping care entirely. The customer needed a fourth option, fast, expert physician care delivered asynchronously, around the clock, on a low flat fee, while meeting the same HIPAA, audit, and clinical-safety standards as a brick-and-mortar practice.
What we built
An AI-assisted, physician-finalized telehealth platform on AWS. Patients describe symptoms in a guided chat conversation; Amazon Bedrock prepares a structured clinical assessment grounded in the practicing physician's own clinical knowledge via Amazon Bedrock Knowledge Bases; Amazon Bedrock Guardrails enforce clinical-safety boundaries throughout the AI workflow; a licensed physician reviews and finalizes the treatment plan; the prescription transmits directly to a pharmacy near the patient, together with any work or school absence notes. Full cycle from symptom intake to transmitted prescription: minutes, not hours. Available 24/7.
Why this matters
The AI's clinical reasoning is grounded in the practicing physician's own knowledge, effectively a digital twin of their diagnostic patterns. The platform extends to new conditions and updated treatment guidelines by editing the underlying knowledge base; no model retraining required. A meaningful share of patients are told upfront they don't need a prescription, a publishable trust signal that the system errs on the side of clinical safety.