Golden Ox

• 

• 

Inspiration

Emergency medicine is stressful — especially when you’re a new student trying to save lives and remember your protocols! We wanted to create a way for EMTs to practice critical decision-making safely, using real-world patient scenarios without the real-world pressure. Golden Ox was born to make EMT training interactive, engaging, and accessible to everyone who wants to master lifesaving care.

What it does

Golden Ox transforms real emergency patient care reports into fun, interactive simulations. Trainees work through lifelike emergencies step-by-step, deciding how to assess, treat, and transport patients. The AI guides them based on their actions, offering hints, corrections, or congratulations — just like a real instructor would (but cuter). 🐂

How we built it

PDF Parsing: We built a pipeline to extract key information from Maryland’s EMEDS patient care reports.

Scenario Generation: We standardize the extracted data into JSON and feed it into a custom prompt to GPT-4 Turbo.

Interactive Simulation: Users interact with the AI through free-text responses, while strict JSON rules manage the simulation flow step-by-step.

Frontend: Built with Next.js, featuring drag-and-drop PDF uploading and a playful UI to keep the experience cozy but serious.

Challenges we ran into

Parsing messy real-world PDFs was harder than expected — especially when formats changed between reports.

We needed the AI to strictly follow EMT protocols (not just improvise), which meant a LOT of careful prompt engineering and system design.

Managing state between user inputs, GPT outputs, and simulation progression without leaks or bugs took several redesigns.

Making it easy and fast for students to start a scenario with their own data (no manual editing required) was tricky.

Accomplishments that we're proud of

We turned chaotic real-world EMT reports into a smooth, immersive educational game.

Built a working AI simulation engine that enforces correct learning steps (like scene safety first!) without "spoiling" answers.

Designed a simple drag-and-drop upload that lets users generate new training scenarios instantly.

Made emergency care training feel safe, low-stakes, and fun while still being medically accurate.

What we learned

Real-world data is messy, but with creativity, you can turn it into something beautiful.

Prompt engineering is an art and a science — small wording changes massively affect AI behavior.

Users love interactivity and feedback — small encouragements ("Nice work!" or "Almost there!") make learning feel motivating.

Good educational tools need to balance realism with fun to keep users engaged.

What's next for Golden Ox

More realistic visuals: Generating cozy, doodle-style scenes based on the patient’s narrative summary.

Dynamic multiplayer mode: Students can team up — one assessing airway, another managing transport.

Student feedback: Provide feedback to students on their performance and things to remember on the national exam

Mobile version: Train anywhere, even from an ambulance bay or your living room couch. 🛋️🚑

Built With

• javascript
• next
• openai
• pdf-parser
• react