InsuraIQ - Your virtual health insurance agent

Health insurance is broken. We analyze your health profile, recommend transparent policies with clear pros & cons, and put you in control. No jargon, no surprises.

Comment

Your Personal Health Insurance Navigator

What Inspired Us

It started with a friend's brother. He was recently diagnosed with a rare hormone disorder that requires a daily injection of medicine, likely for the next decade. The cost? About $35 per shot. $35 a day doesn't sound like a catastrophe, until you do the math. That's over $12,000 a year. We watched in helpless frustration as his family tried to find a health insurance policy that would cover the treatment. They were tossed into a labyrinth of dense policy documents, confusing jargon, and customer service dead-ends. Their story, while deeply personal, is tragically universal. So many of us are forced to make critical financial and health decisions in the dark. We wade through a sea of options, terrified of a "gotcha" clause buried on page 47, only to be quoted an unaffordable premium or get rejected outright because of our medical history. The process feels intentionally broken. We realized the problem wasn't a lack of options, but a complete lack of clarity and advocacy for the user. We couldn't stand by and watch people gamble with their health and finances. we knew there had to be a better way—so we decided to build one.

How We Built It:

To solve a problem this complex, we couldn't just build a simple app. We needed to build a team. So, we used a Multi-Agent System, where specialized AI agents collaborate to guide you through the process, all powered by Google Cloud's robust infrastructure like Agent Engine and Cloud Run.

  • The Opaque Information Collector: Privacy is our absolute priority. This agent is the only one that ever sees your Personally Identifiable Information (PII). It lives in a completely separate, ultra-secure "digital vault" (an A2A server).

  • The Doctor Agent: Once the user’s data is securely handled, this agent analyzes their anonymized medical history. It doesn't know who the user is, only what their health needs are. It then creates a risk profile, identifying potential red flags and key considerations for insurers.

  • The Policy Recommender: This agent takes the anonymous risk profile from the Doctor Agent and goes to work. It communicates with our policy database using MCP, searching for the best possible plans that align with your specific needs and profile.

The "secret sauce" tying this all together is the combination of the Agent Development Kit (ADK), which gives us the flexibility to build this sophisticated agent team, and Google's Vertex AI Gemini, the state-of-the-art intelligence that allows our agents to handle complex requests with human-like reasoning.

What We Learned: Key "Aha!" Moments

Building on the cutting edge means learning on the fly. Here are our biggest breakthroughs:

  1. The Devil is in the Details (of Security): We learned that a secure vault is useless if you can't get information out of it safely. Using the the callback functionality of ADK framework, We made sure only the A2A can see the users sensitive information.

  2. Choosing the Right Brain for the Job: We originally planned to use Gemma, a powerful open-source model, as the core intelligence for our secure agent. While capable, we found its reasoning struggled with the complexity of the logic. We made the call to upgrade to the enterprise-grade Vertex AI Gemini Flash 2.5. It not only provided the sophisticated reasoning we needed but also came with Google's enterprise security guarantees, ensuring user data is never used for training.

  3. Automation Isn't Enough; You Need Empathy: A simple, rigid questionnaire just doesn't work. People's health stories are unique and nuanced. This reinforced our belief that this problem can't be solved with basic automation; it’s a challenge tailor-made for goal-driven, Agentic AI that can adapt to each individual user.

Challenges We Faced:

  1. The A2A ADK Quagmire: The single greatest challenge was structuring the A2A ADK adapter in a way so that we could communicate with the A2A server like your normal ADK LLM agent.

  2. Earning Your Trust, Not Just Asking for It: Asking for medical history is a huge responsibility. Our challenge was to design a system that not only is secure but feels secure. The complete opaqueness of the A2A server, combined with custom logic preventing any data leakage, became our most powerful tool for building that trust.

  3. Taming the Creative Genius: This by far is the biggest challenge we faced. The main volatility of the solution is how the LLM presents information to the user which leads to the unordered way of rendering the questionnaire form. More prompting for us yaay (-‿-")!

What's Next for InsuraIQ

This is just the beginning. Our vision is a comprehensive solution that truly puts the user in control of their health journey. Here's what's on the horizon:

  • Expanding the Library: Growing our policy database to be the most comprehensive resource available.
  • Integrating Public Options: Adding government schemes and public health programs so you can see every single option available to you.
  • Effortless Onboarding: Allowing you to simply upload unstructured data like medical reports and letting our AI build your health profile for you.
  • Closing the Loop: Creating a seamless connection from our platform directly to an insurance agent for the policy of your choice, making the final step as easy as the first.

Built With

Share this project:

Updates