MediClear

Inspiration

Jaina has spent hundreds of hours volunteering at hospitals across Northern California, watching the same situations arise over and over again. Whether she was interacting with patients and families at front desk, or with patients in the nurse's department, she saw firsthand how language often proved to be a barrier in conveying important information to patients and their families. Immediately after the doctor delivers a diagnosis, there is often silence, as patients sit and try to comprehend the gravity of their prognosis. Providing them with the information they need early on to make calculated medical decisions can not only improve morale but save lives.

Our specific interests in this project were also based on personal experiences. Keryssa has been the patient, left confused and feeling out of the loop, unable to make decisions regarding her own body. Having previously undergone surgery, Keryssa has felt firsthand the helplessness of not knowing what was being done to her body.

Jaina has also been on this side of the curtain, having been only six when she had to undergo her first medical procedure: her first EKG to check for any heart abnormalities she had inherited from my father, who passed away at thirty from cardiac arrest. Terrified of the wires strapped to her chest, Jaina listened as my doctor calmly explained how to read her heartbeat, turning fear into fascination. This experience shaped her perspective on medicine – not only as treatment, but as connection. As an aspiring pediatrician, her inspiration stems from this dedication to patient empowerment, advocating for children not only through compassionate care, but by providing them with all the information they need to contribute to their treatment.

What it does

Our project was built to work towards solving this language and literacy barrier, specifically by translating complex medical terminology into simple language for patients to better understand their diagnoses and prognoses. The target audience for users is doctors who can choose an audience mode - "Kids", "Adults", or the "Elderly" - and receive a simple explanation, analogy, and a reassurance for what the future would look like for the patient. The read-aloud feature is for patients to hear their prognosis, for those who are unable to read, while the text-to-text feature is for diagnoses to be sent out to the patients after their appointments are over. Translations can also be saved to a personal glossary to reference after appointments and send to family/friends.

How we built it

MediClear is a React web application built with Vite and styled with Tailwind CSS. The core translation engine is powered by the Google Gemini API, which we prompt-engineered to return structured, audience-calibrated explanations, adjusting vocabulary, tone, and reassurance level depending on whether the patient is a child, an adult, or an elderly person. A MongoDB Atlas database stores a pre-seeded glossary of medical terminology. A Node.js/Express backend handles API proxying to keep credentials secure.

Challenges we ran into

The hardest problem wasn't technical — it was linguistic. Prompt engineering Gemini to reliably produce responses that were clinically grounded but genuinely accessible, without being condescending or alarming, required far more iteration than we expected. Calibrating the difference between an explanation for a ten-year-old and one for an eighty-year-old — in tone, pacing, and word choice — is a genuinely difficult design problem, and one we are still refining.

On the technical side, securely managing API credentials across a Vite frontend and Vercel serverless backend was new territory. We also had to make hard scoping decisions early — staying focused on oncology rather than expanding to all of medicine, and committing to three audience modes rather than an open-ended slider — in order to ship something complete and usable rather than something ambitious and broken.

Accomplishments that we're proud of

We both come from different academic backgrounds, with Keryssa having studied purely computer science and data science, and Jaina having a beginner-level background in data science and a primary focus on biological research and healthcare. We are proud of how we combined our skill sets to bring forth the final product, while also honing in on skills we needed to improve. Keryssa learned more about the backend of healthcare, developing a genuine understanding of what patients and clinicians could need, while Jaina learned how to work with APIs and translate her background in biology to make product decisions like how the glossary was written and how explanations are conveyed for each audience.

What we learned

This is just the beginning. The hardest part is deciding how to practically execute our idea. The most important thing was communication, deciding what to say, how to say it, and who you are saying it to. This was true for us as a team, but will also be true for doctors talking to patients.

What's next for MediClear

Currently, we have incorporated a database that contains a variety of medical procedures with definitions and analogies targeted towards different age groups. Next, we hope to expand across the language barrier, not only translating complex medical language into simpler terms, but translating English to other languages. The website is better conditioned for doctors, such as family medicine practitioners and internal medicine physicians to use during patient appointments, but could eventually be converted to an app that utilizes the voice to voice feature. The aforementioned would be better used by volunteers or nurses conversing with patients over the phone. Lastly, we hope to incorporate diagrams to showcase what certain diagnoses physically look like in the body, to better explain for visual learners.

Built With

• css
• elevenlabs
• figma
• gemini
• javascript
• jsx
• mongodb
• postcss
• react
• tailwind
• vite
• webspeech