StudyMed

StudyMed is a platform for medical students that brings together international research collaboration, interactive courses, and clinical simulations.

Comment

Inspiration

While studying and helping fellow medical students, we noticed that exam and career preparation is scattered across dozens of disconnected sources: textbooks, separate quizzes, forums, ECG practice sheets, standalone writing tools, and random Discord groups for finding research co-authors. Every stage of learning, from interpreting a test result, to writing a research paper, to preparing for clinical decisions, required a different tool, a different login, a different context.

This fragmentation isn't just inconvenient. In medicine, a lack of confidence in fundamental skills such as reading an ECG, interpreting a result, or structured clinical thinking directly affects how prepared a student feels for clinical practice or an internship.

We wanted to build one place designed around how medical learning actually happens, not another quiz app, but a platform for training real skills. What started as a simple idea for an ECG learning game grew into a broader educational platform covering clinical training, academic writing, and research collaboration.

What it does

StudyMed gives medical students one place to learn, practice, and write, starting from the first contact with the material all the way to being ready for real clinical decisions.

Core features include:

Free, accessible topic based courses. Interactive learning games such as ECG Reader and Lab Reader, with timed cases, scoring, and structured feedback after every attempt. YourClinic, a virtual clinical decision making simulation. Research Partners, a space to find co authors and research collaborators from around the world. YourPaper, an academic writing workspace with citation, structure, and quality suggestions. MedGenius, which helps students understand difficult topics in a simpler way.

As you can see, this is a full path: understanding a topic through courses, applying that knowledge through games and the virtual clinic, going deeper by finding research partners, and finally writing original academic papers.

At every stage, MedGenius, powered by GPT 5.6, supports understanding through a novel context based method. A student highlights the part of the text they don't understand, and that highlighted fragment is sent to the AI agent together with the surrounding page content for context, enabling a clearer and simpler explanation tailored to exactly what's confusing.

The platform is built strictly for educational purposes. At no stage does it provide diagnoses or operate on real patient data.

How we built it

We used GPT 5.6 early on to help choose the right technology stack and to shape how we communicated with Codex. Working through single, focused prompts made our process systematic and structured, and that shows in the results.

StudyMed is a web application built with Next.js, React, TypeScript, Tailwind CSS, and Supabase, deployed on Vercel.

The database and authentication logic power user accounts, game progress, demo data, and course content. The project structure was built modularly, so each feature, including games, YourPaper, YourClinic, MedGenius, and Research Partners, could grow independently while sharing the same account system and the same interface language.

We used GPT 5.6 and Codex as real engineering collaborators throughout the build. We thought through the problem and product priorities, while Codex helped us understand our own evolving codebase as the project grew, turn product needs into concrete implementation plans, build and refine frontend components and backend logic, catch edge cases we hadn't anticipated, debug issues using logs and targeted tests, and prepare the project for release through configuration, testing, and documentation.

We never treated AI generated code as automatically correct. Every feature followed the same process: define the problem, review the proposed solution, test it locally, verify the full workflow, and only then consider it done.

Challenges we ran into

One of the biggest challenges was that the ECG Reader game worked great locally, but user scores and progress would sometimes get lost after refreshing the page, and this was exactly the feature we wanted to showcase as the platform's strongest highlight. We broke the problem down into concrete hypotheses together with GPT 5.6, considering whether it was component state, a Supabase write issue, or client side sync, and turned that into a precise prompt for Codex, which located the exact spot in the code and fixed it without touching the rest of the game.

Another challenge was extending the platform without breaking what already worked. Every new feature had to share the same accounts, permissions, and interface patterns instead of becoming another disconnected tool. For example, completing a course through MedGenius should naturally connect to the progress visible in a student's profile.

Data security was equally important. The platform touches educational data and user accounts, so secrets and credentials had to stay out of the repository, and access to features had to be properly restricted by role.

Finally, there was the classic works on my machine moment. A folder named Course ran fine locally on macOS but could have broken production on Vercel, where the file system is case sensitive. We caught this during a repository audit right before publishing and fixed it before anyone ever saw it.

Accomplishments that we're proud of

We're proud that StudyMed grew from a simple educational game idea into a platform covering practical training, academic writing, clinical simulation, and research collaboration, with real room to keep growing.

We're also proud that the project demonstrates what's possible when real understanding of a medical student's needs is combined with effective AI collaboration. It allowed us to turn recurring learning problems into a tool that genuinely makes skill building easier. Understanding how GPT and Codex work together in this way could pay off well beyond this project.

What we learned

The biggest lesson was that small, precise prompts consistently outperformed one giant do it all request. We learned to work effectively with Codex by breaking problems into stages, first clearly understanding what's actually happening, then a narrow, specific task to execute, then verifying the result, and repeating.

GPT 5.6 helped us with the first part, thinking through and framing the problem so that the prompt for Codex was specific rather than vague. Codex excelled at the second part, executing exactly what we asked for, quickly and precisely. This combination consistently outperformed every other approach we tried.

We also learned that the quality of AI output depends heavily on the clarity of the context we provide. The more precisely we described the educational problem, the better the result.

What's next for StudyMed

Next, we want to bring the platform to a full production version and start building partnerships with medical universities to help drive StudyMed's continued development.

We see this as a natural next step. Universities have students who genuinely need a tool like this, along with the subject matter expertise to help us grow educational content responsibly and reliably. We want to build StudyMed not just for students, but together with universities, so the platform truly reflects what medical education needs, not just what we assume it needs.

We believe this platform can genuinely help many students and contribute to more effective learning.

Built With

Share this project:

Updates