Inspiration
Hospitals often experience delays between insurance approval and surgical inventory preparation. Even after coverage is verified, manual communication between billing departments and inventory teams can slow down procedures and increase operational friction. We wanted to build a system that eliminates this gap using AI agents that work together in real time.
PulseLink AI was inspired by the idea of creating an AI workforce (not just a chatbot) that automates high-friction administrative workflows in healthcare.
What it does
PulseLink AI is a multi-agent workforce built using Google’s Agent Development Kit (ADK).
It connects two critical hospital workflows:
Agent 1: Billing & Insurance
- Extracts and interprets insurance coverage data from member cards and policy documents using document parsing.
- Determines coverage status and identifies potential pre-authorization requirements.
Agent 2: Inventory Manager
- Verifies real-time availability of surgical kits and medications.
- Reserves required items once insurance approval is confirmed.
Together, these agents automate the workflow from insurance validation to surgical inventory reservation, reducing delays and manual coordination.
How we built it
We built PulseLink AI using a modern AI agent architecture:
- Orchestration: Google Agent Development Kit (ADK)
- Infrastructure: Google Cloud Vertex AI Agent Engine (managed session state)
- Intelligence: Gemini 2.5 Flash for low-latency reasoning
- Frontend: React-based dashboard with role-based interfaces (Patient, Admin & Doctor)
- Document Processing: Integrated document parsing for insurance extraction
The system is designed with modular agents that communicate through orchestrated workflows, making it scalable and production-ready.
Challenges we ran into
- Designing a clean multi-agent architecture within a limited hackathon timeframe.
- We were initially doing a third agent to automate booking but we did not have time.
- Handling authentication and cloud infrastructure setup.
- Managing real-time session state between agents.
- Designing a UI that feels minimalistic yet futuristic for the Peraton UI/UX track.
- Ensuring our document parsing logic aligned with healthcare use cases.
Despite these challenges, we successfully delivered a working multi-agent workflow prototype.
Accomplishments that we're proud of
- Successfully implemented a functional two-agent workflow.
- Built a clean, intuitive dashboard for admin, patient and doctor views.
- Integrated document parsing into a healthcare workflow.
- Leveraged Google ADK and Vertex AI for real-world agent orchestration.
- Designed a solution that directly improves hospital operational efficiency.
What we learned
Through this project, we learned:
- How to design and orchestrate multi-agent systems using Google ADK.
- The importance of session state management in AI-driven workflows.
- How low-latency models like Gemini 2.5 Flash improve real-time decision making.
- How AI agents can move beyond chatbots into operational automation tools.
We also gained experience designing systems that align with real healthcare infrastructure constraints.
What's next for PulseLink AI
Possibly and hopefully:
- Full EHR system integration.
- Live hospital inventory API integration.
- Predictive surgical supply forecasting.
- Multi-hospital deployment scaling.
- Audit trail and compliance logging for HIPAA-aligned environments.
PulseLink AI can have the potential to become a scalable AI workforce for hospital operations nationwide.
Built With
- gemini-api
- google-cloud-sql
- node.js
- postgresql
- python
- react
- vertex-ai
Log in or sign up for Devpost to join the conversation.