Kaizen Stat
Pocket Swasth — Offline Clinical AI Companion & Patient Twin Simulator
"Using Stat For Better Change"
Submitted
May 22, 2026, 7:11 AM
Last Updated
May 22, 2026, 7:11 AM
Project Links
Problem Statement
Modern healthcare is critically broken for over 70% of low-resource and rural populations. Existing digital health tools are built exclusively for connected urban hubs, failing completely in internet-depleted regions. Furthermore, millions face an insurmountable clinical communication barrier due to language diversity. This lack of access is compounded by a systemic trust deficit—fueled by the rise of counterfeit prescriptions, unverified medical reports, and expensive wearable diagnostics that are out of reach for average families. For doctors, clinical treatment is reactive and speculative; they lack safe, real-time sandboxes to simulate drug efficacy or patient vital trajectories before writing a prescription. These challenges lead to delayed diagnoses, elevated mortality rates, and highly fragmented, untrustworthy patient care.
Solution & AI Usage
Pocket Swasth is a revolutionary hybrid clinical companion operating seamlessly both online and offline. Multilingual Edge AI Triage: Natively compiles and runs lightweight, quantized clinical LLMs (Qwen 0.5B GGUF via on-device compilers) providing fully offline, guardrailed symptom triage and guidance across 10 major Indian languages without internet latency or privacy leaks. Physiological Digital Twin Sandbox: Empowers doctors to instantiate patient-specific digital twins to run high-fidelity 'Hit & Trial' drug trial simulations—predicting 15-day vital shifts, therapeutic efficacy, and contraindication risks powered by advanced Cloud AI (Gemini Flash) with automated local heuristics failovers. Autonomous Verification & Telemetry: Uses on-device OCR for prescription fraud detection, alongside camera-based rPPG algorithms to capture vital signs (heart rate/stress) without costly wearables. Optimized in Flutter for low-end hardware, it makes predictive, localized healthcare a universal
Full Description
Project Description: Pocket Swasth Pocket Swasth is a decentralized, offline-capable clinical assistant and patient digital twin ecosystem designed to bring high-quality healthcare to low-resource and language-diverse populations.
By utilizing advanced edge AI and smart diagnostic tooling, it turns any basic smartphone into a secure, private medical assistant—operating 100% offline without relying on cellular internet.
🌟 Key Features & AI Innovation On-Device Multilingual Edge AI (Offline LLM) Natively compiles and runs a quantized Qwen 0.5B Chat LLM directly on mobile hardware. Provides medical triage and clinical guidance in 10 major Indian languages (Hindi, Bengali, Marathi, Telugu, Tamil, etc.) with strict bilingual guardrails that block non-medical queries automatically. Physiological Digital Twin Sandbox A dedicated Doctor Portal enables physicians to securely sign in and instantiate patient-specific virtual digital twins. Doctors run 'Hit & Trial' drug simulations to predict 15-day vital shifts, therapeutic efficacy, and toxicity risks. The system orchestrates cloud-based LLMs (Gemini-2.5-Flash) for real-time predictions and fails over to local heuristics when offline. Computer Vision & Vital Telemetry (No Wearables) rPPG Vital Extraction: Evaluates subtle facial skin color variations via the smartphone camera to estimate heart rate and stress levels without physical sensors. CV Prescription Verification: Uses on-device OCR to read prescriptions, cross-checking medication registries to detect fraud and double-booking errors. Cloud-Synced RAG Database Seamlessly synchronization of indexed local clinical databases when connected, ensuring patients always have the latest certified medical knowledge offline. 🛠️ Technical Stack Frontend: Flutter (utilizing hardware-accelerated Impeller graphics for high-fidelity holographic visual effects). On-Device LLM: Llama.dart FFI bindings loading quantized GGUF models on CPU. Cloud Predictors: OpenRouter API with automated failovers and model attribution tracking. Pocket Swasth bridges the gap in rural healthcare, moving diagnostics from expensive hospitals and cloud servers directly to the edge, making predictive medicine universally accessible.
Tech Stack
Screenshots
5 attached
