The problem
Prior authorization can involve fragmented clinical information, payer requirements, medication data, documentation, and repetitive administrative work. The challenge was to reduce operational friction without treating AI output as an authoritative clinical or submission decision.
The solution
I helped build Paxei as a human-review-first prior-authorization assistant that brings structured workflows, clinical rules, payer information, document generation, and AI capabilities into a single experience — using OpenAI for narrative generation, extraction, audio, image, batch, and scoring workflows.
My role
My contribution centered on turning complex healthcare workflows into usable product experiences: thinking through AI use cases and guardrails, defining workflows, usability, and product requirements, and working across technical and clinical constraints.
Building AI where trust matters
Healthcare AI can't simply be impressive — it has to know its boundaries.
Paxei was designed around a human-review model. The browser is not treated as authoritative for clinical or submission decisions; backend services revalidate authentication, organization ownership, clinical requirements, payloads, document readiness, and transport safety.
AI + Guardrails
AI accelerates workflows while keeping appropriate human review and validation in the loop.
PHI Protection
Sensitive clinical and patient information is protected through controls such as log redaction and encrypted sensitive database fields.
EHR Integration
The architecture supports interoperability through integrations including Athenahealth, eClinicalWorks/Redox, and FHIR/Da Vinci workflows.
Workflow Automation
Asynchronous processing covers areas such as fax dispatch, FHIR polling, EHR write-back, and formulary synchronization.
Outcome
From complex healthcare workflows → one structured, reviewable experience.
