Inspiration
My older brothers are medical students, and every day I see them overwhelmed by the amount of material they need to study. I just wanted more time with them. But they’re constantly buried in books, PDFs, and lecture notes. That made me realize how broken and exhausting the learning process is in med school.
I may not be a med student myself, but I live this problem up close, and that gave me the drive to build a solution. I’ve interviewed dozens of students who feel the same way: they’re smart, hardworking, but they’re drowning. MapMed is about giving them time back, helping them learn smarter, not harder.
What it does
MapMed is building an AI-powered study assistant that helps medical students save time and learn more effectively. It simplifies complex topics, generates high-yield questions, and uses active recall and spaced repetition to boost retention. MapMed frees up hours each week, helping students focus on what truly matters: mastering medicine without burning out. Our goal is to become the go-to AI copilot for med students, from day one of med school to the licensing exams.
How we built it
We’re using a lightweight and efficient stack focused on fast iteration. Our frontend is built with React and Next.js, and we use Firebase (including Firebase Studio) for authentication, database, and rapid prototyping. For payments, we’ve integrated Stripe. We also use TRAE IDE to help us structure the product and UX flows with precision.
On the AI side, we’re using Gemini 3 Flash as our main model, integrated with a custom RAG pipeline that leverages the Google Programmable Search API to retrieve relevant medical content based on user queries. The model then synthesizes this into personalized study material. This setup allows us to move fast, stay flexible, keep costs low, and focus on delivering real value through AI-powered medical learning.
Challenges we ran into
Building MapMed forced us to confront both technical and human challenges early on.
On the technical side, medical content is dense, high-stakes, and context-sensitive. A generic LLM response is not enough: hallucinations, shallow explanations, or low-yield summaries would completely break trust. Designing a RAG pipeline that retrieves reliable, relevant, and exam-oriented content while staying fast and affordable required multiple iterations. We had to balance latency, cost, and depth of explanations carefully.
Another challenge was avoiding “AI noise.” Many students told us they were already overwhelmed by tools that promise productivity but add more cognitive load. We had to be ruthless about simplicity: every feature had to save time immediately, not just look impressive.
On the human side, we faced the challenge of building for a community we’re not formally part of. As a non-med founder, we had to over-index on listening. That meant dozens of interviews, constant feedback loops, and being comfortable throwing away ideas when students told us, bluntly, “this wouldn’t help me during exam week.”
Accomplishments that we're proud of
We’re proud that MapMed didn’t start as a shiny AI demo, it started as a real pain felt daily.
We interviewed dozens of medical students across different years and schools, and consistently validated that time pressure and inefficient studying are universal problems.
Early users told us that MapMed helps them study with intention, not panic, and that it feels like “having a smart upper-year student guiding you.”
Most importantly, we stayed lean. With a lightweight stack and efficient models, we proved we can deliver meaningful AI value without burning cash.
What we learned
We learned that learning is not an information problem, it’s a prioritization problem.
Medical students don’t need more content. They need help answering: What actually matters? What am I likely to be tested on? What can I safely ignore right now?
We also learned that trust is everything in education. Students don’t want magic, they want clarity, structure, and consistency. AI works best when it acts as a thinking partner, not a replacement for effort.
Finally, we learned that being close to the problem beats having the perfect background. Living alongside med students gave us an unfair advantage: we understand the emotional cost of inefficient learning, not just the academic one.
What's next for MapMed
Next, we’re focused on turning MapMed from a helpful tool into a daily habit.
- Expanding our active recall and spaced repetition system to adapt dynamically to each student’s weaknesses.
- Deepening personalization based on year of study, exam type, and specialty interests.
- Running controlled pilots with medical student cohorts to measure time saved, retention, and exam performance.
- Long term, we aim to support students from their first anatomy class all the way to licensing exams, becoming the default AI copilot for medical education.
Log in or sign up for Devpost to join the conversation.