Med-Clear

• 

• 

  Problem

• 

  Parameters

• 

  Solution

Inspiration

With many of us having elderly relatives who navigate the healthcare system's challenges, we sought a solution to expedite the process of receiving the proper care one needs. Through thorough investigation, we saw a huge bottleneck in the pre-authorization of care, as providers needed countless submissions and resubmissions to make sure the planned procedure, supplements, or care was covered by the insurance. With the United States boasting the highest cost of care in the world, and approval times ranging from days to weeks, our product would not only save lives but save providers and patients thousands.

What it does

Reads Insurance Policies: Turns complicated insurance rules into a simple checklist that providers can actually understand.

Reviews patient records: Pulls out key information from doctor notes, imaging reports, and therapy history to see if the requirements are already met.

Highlights what’s missing: If something important isn’t documented (like six weeks of physical therapy), Med-Clear flags it and asks the clinic to upload or confirm the missing proof.

Builds a complete packet: Automatically creates the packet insurers require, written in the exact language they expect, with supporting evidence attached.

Saves time for providers: Instead of spending hours cross-checking documents, the clinic gets a ready-to-submit package in minutes.

Submits and tracks requests: Sends the pre-authorization to the insurer, then shows status updates: pending, approved, or denied.

Tech Stack

Amazon Bedrock: powered the generative AI models that transformed insurance policies into checklists and reviewed patient records

Amazon S3: provided secure, scalable storage for policy documents, patient data, and pre-authorization packets

Python: handled backend logic, prompt engineering, and integration between AWS services

React: for building a simple front-end interface to interact with the system

GitHub: version control and collaboration for our codebase

Challenges we ran into

The largest challenge we faced was building within the AWS ecosystem. None of our members had prior experience with these services, so we had to learn Bedrock and S3 on the fly. Our most technically difficult hurdle was ensuring that the application received and stored both the policy and medical notes, and thereafter, having Bedrock outputs align with the strict and varied insurance policy requirements.

Accomplishments that we're proud of

The accomplishment we are most proud of is developing a working proof of concept that can read policies, review patient records, and provide instant feedback. Despite starting with zero AWS experience, we successfully integrated Bedrock and S3 into a functioning pipeline that providers could realistically use. We’re also proud of how quickly our team adapted to the healthcare problem space, turning a real-world bottleneck into a solution that could save providers hours of manual work and reduce patient waiting times for care.

What we learned

On the technical side, we learned about RAG pipelines (retrieval-augmented generation), document parsing, and securely managing unstructured data in the cloud.

On the healthcare side, we gained a deeper appreciation for how time-consuming pre-authorization is for providers and how many patients are impacted by delays.

On the teamwork side, we discovered the importance of dividing tasks by specialization: some of us focused on the backend infrastructure, others concentrated on frontend design and implementation, while others worked on prompt engineering, user workflows, and presentation.

What's next for Med-Clear

Although the product was designed for the Hackathon, our team hopes to continue building upon it and scale it for general use. The more policies and health issues our system is exposed to, the better it becomes at identifying missing information, submitting forms, and providing a better care experience. Especially with so many payers and providers communicating via faxed forms, our product is in high demand.

Built With

• amazon-dynamodb
• bedrock
• claude
• github
• python
• react
• s3