EaseYee Care

Inspiration

Everyone has to face the reality of aging. A combination of stigma and cultural norms causes many to delay planning for care when they or their loved ones grow older, and even fewer understand how much they need to spend to maintain their desired lifestyle while aging in place.

Sharing this project’s pain points also encouraged others to open up about their experiences navigating eldercare services. One woman I spoke with at this hackathon shared how her mother’s dementia progressed rapidly, leaving her and her brother responsible for researching the care services the mother required.

Unfortunately, current online services do little to guide or educate users about the different types of service providers (e.g., Skilled Nursing Facilities vs. Assisted Living Facilities vs. Independent Living, or Home Health vs. Home Care). It’s a fragmented system, heavily built on handshake deals and the passing of patients between providers—often not in the best interest of the patient.

What it does

I built this hackathon with a focus on how generative AI can provide value. In this use case, we generate Care Plans based on diagnostic keywords. Care Plans consist of comprehensive medical and non-medical tasks that service providers develop when assessing a patient. Since much of the care is templatized, the idea was to use these care plans as a data source and extract a usable plan to help users:

1. Get an idea of the types of services they might need
   
2. Explore cost-effective alternatives to avoid high-cost or high-usage care
   
3. Receive recommendations for placement or home modifications they may need in the future.

How we built it

I used the serverless framework to host a backend lambda service, accessible via API gateway. The backend service hosts our knowledge source, in this case example generated care plans made by Generative AI. Ideally, this will be either human authored care plans, or past care plans created for patients in order to build a more realistic dataset. For the frontend, React + Vite framework is used and connects to the backend via HTTPS requests.

Challenges we ran into

Ideally, the pipeline for care plan should stream content from the Generative AIuthor. I tried going down the AWS Bedrock route, hoping to be able to tie it to my use case, but I got stuck with syncing issues with the data source, which is supposed to be the mechanism that embeds the text from the S3 bucket datasource. If I had to do it over again, I probably would have brute forced it with some other cloud service like pinecone or had mongodb host the embeddings.

The other challenge was: It was also difficult to garner interest on the idea. Either the product wasn't tackling a sexy industry, or I wasn't representing the problem/solution well; either way, I found it difficult to gather teammates to either abandon their ideas or join out of curiosity.

Accomplishments that we're proud of

I would say my biggest accomplishment was advocating for an industry that many people tend to overlook. While refining the hackathon idea, I spoke with a lot of people, and most were curious about why I’m pursuing work in the eldercare space. The truth is, industries like this need more talent to make a meaningful impact, and advocating for our most vulnerable population is just as important as delivering a great product.

Whats next for EaseYee Care?

I'm looking to build this full time! Currently, I'm doing a lot of user discovery and cold calling to learn more about the industry, but hopefully, we can sign some users and get paying customers one day!

Built With

• amazon-web-services
• lambda
• python
• react
• serverless