Redesigning the Residency Match List with AI
Redesigning the Residency Match List with AI
Summary
More than 40,000 medical students use the National Resident Matching Program (NRMP) to submit their rank order list for residency programs every year. This year, I got to see the process from up close. In an already stressful process, the NRMP did applicants no favors.
I wanted to try my hand at a redesign of this platform and see how I can leverage AI to expedite my design process.
Summary
More than 40,000 medical students use the National Resident Matching Program (NRMP) to submit their rank order list for residency programs every year. This year, I got to see the process from up close. In an already stressful process, the NRMP did applicants no favors.
I wanted to try my hand at a redesign of this platform and see how I can leverage AI to expedite my design process.
Role
Sole Designer
Role
Sole Designer
Collaborators
ChatGPT, Claude, Gemini
Collaborators
ChatGPT, Claude, Gemini
Overview
Overview
Becoming a doctor is no easy process, and the residency application process is overly convoluted, even for some of the smartest students.
Becoming a doctor is no easy process, and the residency application process is overly convoluted, even for some of the smartest students.
To become a doctor, you have to apply and get into medical school, pass your Step 1 exam, go through rotations of all specialties, get good evaluations, do amazing on your Step 2 exam, and apply for residency by sending out potentially hundreds of applications to programs across the nation. You have to have nerves of steel to make it through.
I definitely don’t, but this year, my partner successfully matched into his preferred specialty.
To become a doctor, you have to apply and get into medical school, pass your Step 1 exam, go through rotations of all specialties, get good evaluations, do amazing on your Step 2 exam, and apply for residency by sending out potentially hundreds of applications to programs across the nation. You have to have nerves of steel to make it through.
I definitely don’t, but this year, my partner successfully matched into his preferred specialty.
Match Day 2025 at NJMS!
There’s no question that applying to residency is a stressful process. However, there’s no reason that the actual application process should be an added stressor.
The residency application process consists of three platforms:
Electronic Residency Application Service (ERAS): Where medical students go to apply to residency
Thalamus: Where some programs will send you interview notifications
National Resident Matching Program (NRMP): Where med students make their final rank list for the Match.
The residency application process consists of three platforms:
Electronic Residency Application Service (ERAS): Where medical students go to apply to residency
Thalamus: Where some programs will send you interview notifications
National Resident Matching Program (NRMP): Where med students make their final rank list for the Match.
When my partner was creating his rank list, I didn’t expect the process to be so overwhelming.
When my partner was creating his rank list, I didn’t expect the process to be so overwhelming.
The current state of the NRMP Match List
I watched him open up the 17 page pdf to figure out if he could edit his rank list after he submitted it, only to end up on Reddit looking for confirmation.
Questions about the NRMP certification process even go back 16 years…
Questions about the NRMP certification process even go back 16 years…
A question from February 2009, which is the exact question my partner asked in March 2025
Always excited for a potential redesign, I thought this could be the perfect problem to tackle as a way to experiment with AI.
Always excited for a potential redesign, I thought this could be the perfect problem to tackle as a way to experiment with AI.
How might we reduce stress and cognitive overload during a high-stakes moment by improving the NRMP ranking interface?
How might we reduce stress and cognitive overload during a high-stakes moment by improving the NRMP ranking interface?
Goals
Goals
This redesign was focused on making the ranking process the easiest part of the residency application.
This redesign was focused on making the ranking process the easiest part of the residency application.
Specifically, I focused on these three goals:
Specifically, I focused on these three goals:
Reduce cognitive load
Clearly indicate a path to success, removing any noise that obscures that path
Increase User Confidence
Reduce doubt and have fewer med students walk away unsure if they've done things correctly
Make the process more intuitive
Med students should be able to complete their match list without having to leave the platform and watch a tutorial
The Process
The Process
My AI Collaborators took on various forms
My AI Collaborators took on various forms
Synthetic Users
UX Researcher
User Interview Expert
Visual Designer
UX Designer
Design Critiquer
Front-end Developer
AI was honestly great at coming up with personas.
AI was honestly great at coming up with personas.
Based on my experience hearing stories from friends in med school and watching videos online, it’s obvious that no med student is like another.
Based on my experience hearing stories from friends in med school and watching videos online, it’s obvious that no med student is like another, but I was surprised how AI was able to bring forward some nuance without being explicitly told beforehand.
Personally, I was most impressed by Claude’s assessment of user personas.
Personally, I was most impressed by Claude’s assessment of user personas.
A screenshot of a response from Claude with a single prompt:
"As a UXD, generate personas for medical students applying to residency. I am working on a redesign of the NRMP match list and want to know which user personas to target"
I made my own assessment of the issues with the interface, but also asked AI and med students to determine what were the biggest pain points when coming up with a rank list.
I made my own assessment of the issues with the interface, but also asked AI and med students to determine what were the biggest pain points when coming up with a rank list.
Med Students were more process oriented. The entire concept of “Save” vs. “certifying” was the biggest issue. And they just didn’t like how the website looked,
Med Students were more process oriented. The entire concept of “Save” vs. “certifying” was the biggest issue. And they just didn’t like how the website looked,
What I saw was information overload: lots of actions and information that diluted the important details.
What I saw was information overload: lots of actions and information that diluted the important details.
Gemini, Claude, and ChatGPT all gave similar results, calling out an outdated design, unclear hierarchy, and an overwhelming amount of buttons.
Gemini, Claude, and ChatGPT all gave similar results, calling out an outdated design, unclear hierarchy, and an overwhelming amount of buttons.
Getting some help with wireframes was a quick and easy way to combat empty canvas syndrome.
Getting some help with wireframes was a quick and easy way to combat empty canvas syndrome.
Starting from scratch is always harder for me than editing or improving an existing design. After defining the basic concepts the rank list needed to have (a reorderable list, a clear save button, and key status indicators) and curating my own inspiration deck, Claude and ChatGPT were able to give me a good starting place.
Starting from scratch is always harder for me than editing or improving an existing design. After defining the basic concepts the rank list needed to have (a reorderable list, a clear save button, and key status indicators) and curating my own inspiration deck, Claude and ChatGPT were able to give me a good starting place.
An example from Claude of a redesigned NRMP match list
The Solution
The Solution
A new, streamlined approach to creating a simple, reorderable list.
A new, streamlined approach to creating a simple, reorderable list.
Key features of the new design include:
KPIs indicating status, with green checks to provide more reassurance
A FAQ bar with questions and answers readily available
Card-based reorderable list to make it clear what’s been selected, drawing attention to the location and selected field
Key features of the new design include:
KPIs indicating status, with green checks to provide more reassurance
A FAQ bar with questions and answers readily available
Card-based reorderable list to make it clear what’s been selected, drawing attention to the location and selected field
Reflection
Reflection
It's never too late to reevaluate processes and improve the way we do things.
It's never too late to reevaluate processes and improve the way we do things.
Using AI in the ideation process was a major game-changer. This was just one of my early experiments with various AI tools. What I really saw as the biggest benefit was eliminating the need to start from scratch, unless it’s desired.
Using AI in the ideation process was a major game-changer. This was just one of my early experiments with various AI tools. What I really saw as the biggest benefit was eliminating the need to start from scratch, unless it’s desired.
With the NRMP match list process though, there’s no replacement for seeing or going through the process first hand. If I didn’t have the background I did when I approached this design challenge, it would have been a lot harder for me to assess the quality of AI responses.
With the NRMP match list process though, there’s no replacement for seeing or going through the process first hand. If I didn’t have the background I did when I approached this design challenge, it would have been a lot harder for me to assess the quality of AI responses.
With all said and done, I would love to develop this idea further and do more rigorous testing with med students, and maybe even campaign for a change.
With all said and done, I would love to develop this idea further and do more rigorous testing with med students, and maybe even campaign for a change.