SenseAI — AI-Powered Healthcare Assistant

Delivering intelligent, real-time clinical guidance to frontline healthcare workers in fast-paced, high-density urban environments.

Inspiration

Healthcare systems (NGOs particularly) face a different kind of strain — overcrowded clinics, high patient volumes, time pressure, and fragmented data systems. Frontline workers and clinic staff often need to make rapid decisions with limited time, inconsistent patient histories, and operational overload.

We set out to build a solution that provides instant risk assessment and structured clinical guidance, helping healthcare workers act faster and more confidently — even in high-pressure, resource-constrained urban settings.

What It Does

SenseAI is an AI-powered healthcare assistant designed to support patient assessment and clinical decision-making in urban clinics and community health centers. It supports multiple conditions including:

Maternal Hemorrhage
Tuberculosis Adherence
Pesticide Exposure
Diabetic Foot Ulcers (DFU)

The system captures patient inputs, processes them through specialized AI agents, and delivers:

Output	Description
Risk Score & Level	Quantified severity assessment for quick triage
Clear Explanation	Simple reasoning for informed decision-making
Actionable Checklist	Step-by-step clinical guidance
Missing Data Indicators	Identifies gaps to ensure complete assessment

It works across multiple access points:

Mobile App (Flutter) for on-ground healthcare staff
WhatsApp Chatbot for quick interactions
Voice Call Agent for hands-free usage in busy settings

How We Built It

SenseAI uses a hybrid AI architecture tailored for urban healthcare workflows where speed, scalability, and reliability are critical.

Frontend

Flutter mobile application for healthcare workers in clinics and field visits

Backend

Flask APIs managing condition-specific diagnosis agents

AI Models

Model	Role
DeepSeek	Structured reasoning and decision-making
VisionPro + Custom Segmentation	DFU image analysis
Gemma 2B (quantized)	Offline fallback via `llama.cpp`

Integrations

WhatsApp Cloud API — quick-access chat interface
ElevenLabs — voice-based AI assistant

Each condition is handled by a dedicated agent with structured scoring logic, ensuring consistent and reliable outputs across high patient volumes.

SenseAI Dashboard — Clinic Management Layer (Medo)

While the mobile app supports healthcare workers during patient interactions, SenseAI is complemented by Medo — a browser-based clinic management dashboard designed for urban clinics, hospitals, and NGO networks.

What Medo Adds

In urban setups, where multiple workers handle large patient inflows daily, Medo provides centralized visibility, coordination, and analytics.

Layer	Tool	Users
Point-of-care	Flutter Mobile App	Healthcare Workers
Clinic Management	Medo Web Dashboard	Clinic Admins / Supervisors

What Medo Does

Medo is a mobile-responsive, lightweight dashboard that works without heavy infrastructure, making it suitable for clinics with limited IT support.

Authentication & Clinic Setup

Clinics register with basic details and receive a unique Clinic ID
Role-based login for Admins and Healthcare Workers
Data isolation ensures each clinic operates independently

Worker Management

Admins manage healthcare workers across departments or locations
Worker profiles track:
- Patients handled
- Case submissions
- Disease distribution
- Daily consistency

Patient Entry Tracking

Workers log daily patient cases with:
- Disease type
- Patient count
- Severity distribution (Low / Medium / High / Critical)
Built-in validation ensures accurate reporting

Analytics

Designed for high-volume urban data insights:

Disease trends across time
Daily patient load patterns
Severity distribution
Worker performance comparisons

All analytics support filtering by date range, disease type, and worker.

Reports & Export

Generate and export reports in CSV, Excel, or PDF format
Useful for hospital administration, NGO reporting, and audits

Daily Reminder System

Flags missing daily submissions
Ensures no operational blind spots in busy clinics

Supported Conditions

Maternal Hemorrhage
Tuberculosis
Pesticide Exposure
Diabetic Foot Ulcers

How Medo Complements the Mobile App

The system creates a closed-loop urban healthcare workflow:

Healthcare Worker (Mobile App)
  → Quickly assesses patients and generates risk scores
  → Logs case data into Medo

Clinic Admin (Dashboard)
  → Monitors patient inflow and worker activity
  → Tracks disease patterns across the clinic
  → Generates reports for operational and compliance needs

This ensures real-time action at the patient level + strategic visibility at the clinic level.

Challenges We Ran Into

Handling high patient throughput without slowing down the system
Designing AI outputs that are fast yet reliable under time pressure
Managing online-offline switching in semi-connected urban zones
Ensuring consistency across multiple access channels (app, WhatsApp, voice)
Scaling image-based diagnosis for diverse real-world cases

Accomplishments We Are Proud Of

Built a hybrid AI system that works across connectivity conditions
Enabled multi-channel access for different clinical scenarios
Designed structured clinical scoring for faster triage decisions
Integrated image-based DFU analysis into real workflows
Developed a dual-layer system (field + dashboard) for full operational coverage
Created a solution adaptable to high-density urban healthcare environments

What We Learned

In urban settings, speed is as critical as accuracy
Structured AI outputs outperform generic responses in clinical workflows
UI must be extremely intuitive due to time constraints
Systems must handle data scale and concurrency effectively
Field tools without dashboards limit decision-making at higher levels

What's Next for SenseAI

[ ] Optimize performance for high-load urban environments
[ ] Expand to additional conditions — malaria, anemia, child health
[ ] Integrate telemedicine for real-time doctor support
[ ] Improve image models with more diverse urban datasets
[ ] Add support for regional languages and voice workflows
[ ] Transition to open-source AI models to reduce cost
[ ] Enable real-time sync between mobile app and dashboard
[ ] Add offline data buffering for intermittent connectivity scenarios

Tech Stack

Frontend (Field)   → Flutter (Mobile App)
Frontend (Clinic)  → React.js + Tailwind CSS (Medo Dashboard)
Backend            → Flask (Python)
AI (Online)        → DeepSeek, VisionPro
AI (Offline)       → Gemma 2B via llama.cpp
Messaging          → WhatsApp Cloud API
Voice              → ElevenLabs
Image Analysis     → Gemini-Vision Pro

Built for fast-moving healthcare systems where every second — and every decision — matters.