SwasthAI is an Autonomous Health Intelligence Agent Network that demonstrates how AI agents can coordinate to provide proactive health monitoring and early epidemic detection. This hackathon prototype focuses on showcasing true agentic behaviour through autonomous decision-making, multi-agent coordination, and goal-oriented health surveillance workflows accessible via WhatsApp messaging.
The system transforms traditional reactive health reporting into a proactive, agent-driven surveillance network where specialized AI agents autonomously assess symptoms, detect patterns, coordinate responses, and escalate alerts without human intervention.

Key Features

• Agentic Surveillance Orchestration: A central coordinator manages specialized agents for symptom triage, pattern detection, and alerting, producing predictive insights and automated notifications from community-reported data.

• WhatsApp Health Assistant: Real-time symptom assessment over WhatsApp with adaptive, agent-driven dialogues and autonomous follow-up scheduling for check-ins and reminders.

• Pattern Detection & Alerts: Lightweight statistical anomaly detection flags unusual symptom spikes or clusters, triggering agent-led community alerts when thresholds are crossed.

• Mock Government Interop: Simulated connections to health databases via mock endpoints, demonstrating agent decisions for data sharing with basic consent and audit logging.

Technical Implementation

1. Multi-Agent Framework: Orchestrate autonomous agents with a lightweight coordinator for triage, pattern analysis, communications, and alerting in a single end-to-end workflow.

2. LLM Tooling: Use function-calling to perform symptom triage, read/write database records, dispatch alerts, and trigger follow-ups directly from agent decisions.

3. System Integration: Handle real-time WhatsApp messaging via webhooks, route agent tasks through a FastAPI backend, and expose mock endpoints to simulate external health systems.

4. Anomaly Detection: Apply moving-window counts and simple thresholds to flag symptom/location spikes, emitting escalation signals that agents act on automatically.

5. Scalable Architecture: Start with a single-process app and a lightweight DB; roadmap includes cloud microservices, advanced forecasting models, high-throughput pipelines, hardened security, and real integrations.

Specific Pain Points Addressed

1. Early Detection Through Automation

• Autonomous Surveillance: Agents continuously monitor without human oversight.
• 24/7 Pattern Recognition: Automated detection of symptom clusters and unusual health trends.
• Instant Escalation: Agent-driven alerts eliminate human delay in critical situations.

2. Coordinated Response Management

• Multi-Agent Coordination: Specialized agents handle different aspects of health surveillance.
• Automated Follow-ups: Agents schedule and manage ongoing patient communications.
• Smart Resource Allocation: Agents prioritize high-risk cases for immediate attention.

3. Accessible Health Intelligence

• WhatsApp Integration: Leverages familiar messaging platform for broad accessibility.
• Autonomous Interviews: Agents conduct structured health assessments independently.
• Proactive Outreach: Agents initiate contact for follow-up care and community alerts.

Target Audience

1. Primary Users:

• Rural and semi-urban citizens (500M+ underserved)
• Community health workers (e.g., ASHA workers)
• State and district health departments
• National Health Mission (NHM) teams

2. Secondary Users:

• NGOs and public health organizations
• Research institutions and data analysts
• Corporates with employee health programs
• International health agencies (e.g., WHO, UNICEF)

Go-to-Market Strategy

• To Partner with state health departments for pilots integrated with frontline worker networks by aligning with National Health Mission and ABDM’s WhatsApp infrastructure.
• Ensure DPDP compliance and clinical validation.
• Collaborate with cloud providers and telemedicine platforms.
• Promote at AI and healthcare forums for credibility and global partnerships.

Revenue Model

• Per Alert Pricing: Charges may range from ₹50 to ₹1,500 per health alert generated, scalable by volume and alert complexity.
• Subscription Plans: Tiered monthly subscriptions starting at ₹1 lakhs for supporting baseline alert volumes, with additional charges for exceeding thresholds.
• Enterprise and Government Licensing: Custom pricing for large-scale clients integrating alert systems into their infrastructure, enabling wide coverage and performance-based incentives.