Cognitive rehab, built for the patient and the people around them.
An iPadOS platform that makes structured cognitive exercises feel less like clinical homework and more like progress.
Neuro Rehab is an AI-orchestrated cognitive rehabilitation platform designed for stroke patients recovering at home. It wraps a pluggable game library, real-time difficulty adjustment, and caregiver visibility into four role-shaped apps that share a single design system — all running offline-first on iPad with Firebase cloud sync.
After a stroke, cognitive rehabilitation is critical — but the tools available are stuck in a clinical setting. Therapists hand patients paper worksheets or use desktop software designed for a lab. Exercises are rigid, feedback is delayed by weeks between appointments, and caregivers have almost no visibility into what’s happening day to day.
The patients who need this the most are often the ones with the least access: people recovering at home, in rural areas, or with limited mobility. They need something they can pick up on an iPad at the kitchen table, something that adapts to how they’re doing that day, and something that keeps their therapist and family in the loop without turning recovery into a reporting burden.
Existing rehab apps tend to be either too gamified (losing clinical validity) or too clinical (losing patient engagement). The challenge was to build a system that is medically grounded, adapts in real time, works without an internet connection, and wraps the patient in a care network that can actually help.
Neuro Rehab is pre-production — the architecture is designed, the domain model is mapped, and the system contracts are specified. The primary visual is the architecture itself, which tells the story of four role-shaped apps built on a shared foundation.
The Patient App is the primary autonomous experience: AI-guided session flow, cognitive exercises drawn from a pluggable game library, and real-time encouragement. The Caregiver App surfaces wellness checks, medication reminders, and activity feeds. The Therapist App provides cohort analytics, treatment plan editing, and score trending. The Family App offers encouragement tools, shared moments, and weekly summaries.
The system is five layers deep, from the SwiftUI presentation layer through application-level orchestration, domain-driven bounded contexts, an offline-first data tier, and a thin Firebase cloud backend. Cross-cutting concerns — resilience, security, HIPAA compliance, privacy, observability, and accessibility — apply vertically across every layer.
Four role-shaped apps, one design system.
Patients, caregivers, therapists, and family members each get a purpose-built app, but they all draw from a shared design system with stroke-safe touch targets, high-contrast tokens, and accessibility-first components.
Offline-first is not optional.
Stroke patients recovering at home cannot depend on reliable connectivity. Every exercise, every score, every session state is persisted locally first and synced to Firebase when the network is available.
AI orchestration with a policy layer.
Not every decision needs a cloud LLM. The AI router evaluates each request against a policy layer that routes between a rule engine (deterministic, fast, free), an on-device model (private, low-latency), and a cloud model (capable, costly).
HIPAA-aware from day one.
Privacy, security, and compliance are cross-cutting concerns that span every layer of the architecture — not bolted on after the fact. Role-based access control, audit logging, data encryption at rest and in transit, and consent management are baked into the system contracts.
The next phase is building the Patient App MVP: the session state machine, two cognitive game plugins (a memory-matching game and a sequencing exercise), and the offline persistence layer. The goal is a working prototype that a therapist can hand to a patient and walk away — the app adapts difficulty, tracks progress, and syncs results when the patient is back online.