Project Description: CliniMate

Purpose & Problem Statement

In Malaysia’s public healthcare system, clinicians are overburdened, spending nearly 50% of their time on administrative tasks. This reliance on fragmented systems, manual data entry, and paper records leads to significant inefficiencies, cognitive overload for staff, and an increased risk of clinical errors.

CliniMate is an AI-powered clinical workflow assistant designed to solve this problem. Its purpose is to automate repetitive administrative work, streamline data capture, and unify the patient journey across all clinical roles — empowering healthcare professionals to focus on patient care, not paperwork.

---

Users & Workflow

CliniMate is a role-based system designed for four key users in a clinical setting, plus the patient:

• Receptionist
  • Registers new or existing patients, automatically generating a queue ticket and notifying the patient via email.
  • A “Call Next” function seamlessly moves the patient to the next stage.
• Nurse
  • Is notified when a patient arrives.
  • Can upload a photo of a vitals monitor, where our AI pipeline uses OCR to extract key readings (BP, HR, Temp, etc.).
  • Rule-based analysis flags abnormalities for immediate attention.
• Doctor
  • Records audio of the patient consultation.
  • System transcribes the conversation, summarizes it, and generates structured SOAP (Subjective, Objective, Assessment, Plan) notes.
  • Doctor can review and edit these notes.
• Pharmacist
  • Receives the patient in their queue with the doctor’s prescription auto-populated.
  • Prepares the medication and notifies the patient for collection.
• Administrator
  • Oversees clinic operations.

Patient Notifications: Throughout the process, patients receive email notifications. At the end, a bilingual (English/Malay) summary of their visit can be sent, improving health literacy.

---

Measurable Impact & Key Results

• Reduced Administrative Time: Up to 40% reduction in time spent on documentation and inter-departmental coordination.
• Improved Data Accuracy: Automation of data entry from vitals monitors and voice notes minimizes human error.
• Faster Patient Throughput: Unified queue and automated notifications ensure smooth patient flow.
• Enhanced Patient Experience: Patients are informed at every step and leave with a clear, digital visit record.

---

User Experience & Scalability

The UX is centered on simplicity and efficiency, with role-specific dashboards showing only relevant information.

• One-click automation for queue management.
• AI-powered data entry reduces cognitive load.
• Built on Supabase for a scalable backend (database, authentication, serverless functions, storage).
• Ready for deployment in real-world clinics like Malaysia’s Klinik Kesihatan.

---

Judging Criteria Analysis

Innovation

CliniMate tightly integrates multiple AI services (OCR, Speech-to-Text, LLM-based summarization) directly into clinical workflows — creating a cohesive, automated system, not just a standalone tool.

Feasibility

A working full-stack application built on mature, accessible technologies. Core features are implemented and functional, ready for pilot deployment.

Technical Merit

Uses a modern tech stack (React, Supabase) and demonstrates advanced AI integration:

• Fallback mechanisms (e.g., browser SpeechRecognition if Whisper fails).
• Combination of AI with deterministic, rule-based logic for safety and reliability.

User Benefit

Gives clinicians back their most valuable resource: time, enabling:

• Better, safer patient care.
• Reduced burnout.
• Improved operational efficiency.

Deployability

As a web-based application on Supabase:

• No local installation required.
• Scalable cloud backend.
• Suitable for outpatient departments or Klinik Kesihatan facilities.

---

Technology Stack

Frameworks & Libraries:

• React, Vite, TypeScript, Tailwind CSS, Shadcn/UI, TanStack Query, Zod, React Hook Form, i18next.

Backend & Cloud Services:

• Supabase (Postgres Database, Auth, Storage, Edge Functions).

APIs:

• Supabase API, OpenAI API, Hugging Face Inference API.

AI Models & Approaches:

• Voice-to-Text: OpenAI Whisper (primary) + browser SpeechRecognition API (fallback).
• OCR for Vitals: Hugging Face TrOCR (microsoft/trocr-base-handwritten).
• Transcript Summarization & SOAP Notes: Hugging Face summarization models (e.g., BART) + fine-tuned LLMs.
• Rule-Based Analysis: Deterministic TypeScript functions for clinical threshold checks (e.g., hypertension, fever).
• Deterministic NLP: Regex & rule-based parsers for reliable OCR/text preprocessing.

---

Future Improvements

• Proactive AI Assistant: “Health Twin” for real-time clinical decision support (e.g., drug interaction alerts, diagnosis suggestions).
• Predictive Health Analytics: ML models for early detection of chronic disease risks.
• Full EMR/EHR Integration: HL7 FHIR-compliant integration with hospital record systems.
• Patient Mobile App & Offline Mode: Appointment management, health record access, and offline-ready clinician app (PWA).