ba.chat

Inspiration

Our project is inspired by the lack of resources to clarify healthcare plans and simplify insurance for students. All new adults undergo a steep learning curve as they are exposed to the world, and healthcare, albeit the most important tool in adulthood, is one of the most overlooked.

Our passion for this cause stems specifically from the story of our friend, who faced a serious leg injury a few months ago and needed operations to recover. Being an international student, he was unaware of the complexity of the American healthcare system. As such, he spent far too much time in his condition bogged up in the details of how to afford his bills, and finds himself even now finding deductibles in his medical bills.

We think that it is unfair to assume that students, especially international students, be aware of the maze of fine print that is American healthcare. Thus was born our brainchild, ba.chat - a play on the Hindi word bachat, meaning savings.

We recognize that your time is money. ba.chat saves both.

What it does

ba.chat provides a host of options to simplify the world of healthcare for users. These are the key functionalities we have implemented into the app right now:

Analyze the cost of treatment and all deductibles based on your plan.
Obtain contact information for the best treatment and hospitals in your vicinity - balanced for a blend of affordability and premium care.
Craft a convincing email to your healthcare provider, underlining terms in your policy that the average user would gloss over. Maximize your benefits!
Check for potential medical error: Medical errors are the third leading cause of death in America, and ba.chat cross-references the depths of medical resources to ensure you are not put in danger.

The application amalgamates Reflex, Gemini API, and Python 3.11 to provide real-time feedback to users in a simple yet effective manner. Apart from the main use cases provided above, users are also free to ask any follow-up questions or general healthcare questions to ba.chat and treat it as their free personal healthcare assistant.

How to use it

Our super-easy-to-use pipeline:

Upload your insurance provider, location, email address, and most recent diagnosis into their designated areas.
Simply click on the service you seek from the options provided: Insurance Analysis, Scheduling Appointments, Negotiating Prices, and Double Checking for Medical Error.
Receive a response tailored to your specifications exactly, and chat with our friendly AI to get more information!
Ask ba.chat to send an email to your healthcare provider to negotiate costs or hospitals near you to schedule appointments!

Challenges we ran into

While we had a great time working together, it wasn't all smooth sailing. Here were some challenges we faced in our process:

New frameworks: Figuring out the entire Reflex framework to build the platform took up a lot of our time, since we had to go through hordes of documentation and sample apps to figure out how the different components of Reflex work together. While there were various templates that already existed, we wanted to make our platform unique - not just have a textbox and chatbot. This is why we integrated different services along with features as well.
Prompt tuning: A major part of this was leveraging the power of the Gemini API by prompt-tuning the model to take in a specific kind of input and give us the required output. This involved a lot of research into optimizing prompts (essentially learning prompt engineering), testing out different prompts, getting incorrect answers with insufficient data or too much text, or just unsatisfactory responses. A big part of prompt tuning was finding that fine balance to get the results that we wanted.
Locally based issues: working with different environments on different devices and coordinating our work was quite challenging initially. While this may seem trivial, we learnt a lot about teamwork and pair programming and it pushed us to work together as a whole instead of working differently. It pushed us all out of our comfort zone and helped us develop a platform as a team.

What we learned

Building ba.chat was an enriching experience through and through. Here are some of our key takeaways:

Teamwork: While this may sound cliche, we genuinely learnt how to work together as a team. We all have programming backgrounds and are well-versed with various tech stacks but have rarely gotten chances to work together to build such a large-scale project in such a short period of time. At first, we were working together well but we had to learn how to navigate the clashing of our ideas and frustrations as we ran into various issues in the project. We all had to compromise and ultimately our project was the result of the best of our ideas coming together.
Going out of our comfort zone technically: We have all had experience with various tech stacks but none of us had ever used the Reflex framework before. This forced us out of our comfort zone as we all had to adapt to a different way of doing things. This ultimately strengthened our flexibility and we learnt that if our development skills and persistence were strong, we could figure out any new framework.

What's next for ba.chat

Scaling:

We have various ideas on how to scale our model so that it can reach a much larger audience:

Expanding to other universities: We aim to expand our services to universities beyond the UCs, so that all students in the United States have simplified healthcare options.
Expanding to other healthcare providers: We hope to integrate options with larger providers such as Kaiser Permanente and Cigna, improving our model to work on their extensive and varied healthcare plans.
Multilingual and all-inclusive options: To drive home the point of what our project is really about, we aim to integrate multilingual capabilities into our platform, and make it more inclusive for various marginalized communities. This will provide healthcare access to those who really need it most.

Improvements:

Additionally, we have various ideas in the works for how to improve our product already:

Integration with Google Maps: Integrating Gemini with Google Maps would provide far more accurate location services, allowing us to provide the perfect blend of convenience, affordability, and quality of care.
Integration with Google Calendar and Mail: This would allow us to make our pipeline truly end-to-end, ensuring that the user does not have to leave our platform for the entirety of their healthcare negotiation and management.
Fine tuning prompts: Better prompts for the APIs under the hood would allow us to improve the quality of the model tenfold, and we think we can get there with a little tailoring.