THE RESIDENT’s CORNER –
NEW to the SEA Newsletter is a section called – “the Resident’s Corner”. Thanks to the residents who have contributed to the first content pieces debuting this effort. We will highlight 1-3 write-ups in each Newsletter showcasing what residents are doing in the “Anesthesia Education” space relating to the theme of that conference.
During this Fall Meeting featuring the theme, “Innovations in Anesthesia Education: Integrating Pedagogy and Technology”, we kick off “the Resident’s Corner” with these 4 pieces ….
(#1).
Augmenting the Scheduler: Using Technology to Facilitate Anesthesia Residency Scheduling
Monica Ha, MD
Mentors: Kevin Weng BS MEng, Onyi Onuoha, MD, Travis Markham MD.
Department of Anesthesiology, University of Texas, Houston
At UTHealth Houston, scheduling more than 120 residents and fellows across three residency programs and five anesthesia subspecialties means coordinating over 3,000 assignments each month. Beyond routine operating room coverage, more than 20 specially designated shifts must be staffed each day while accounting for vacations, conferences, lecture days, rotation requirements, and more than 60 individual requests.
When I became scheduling chief, I was eager to see whether technology could make this process more manageable. Despite previous unsuccessful attempts at automation, I have spent the past seven months working with my husband, a software engineer, to develop a scheduling application.
The first challenge was translating what experienced schedulers know intuitively into explicit rules. Some are absolute, such as honoring post-call days and ensuring appropriate seniority for call shifts. Others are preferences, such as spacing calls apart and distributing undesirable assignments equitably. Using AI-assisted coding tools, we incorporated both types of rules into the application, alongside rotation requirements, staffing minimums, and other scheduling needs.
Our initial approach treated scheduling as an optimization problem. We used an optimizer to generate more than 500 iterations of each monthly schedule. Yet, more computation did not necessarily produce better results. A mathematically favorable solution could satisfy hundreds of constraints while still distributing weekends, shifts, or calls unreasonably. Correcting these schedules sometimes took longer than rebuilding them.
The turning point came when I reconsidered how I scheduled manually. I built each month in layers, establishing the most important assignments first, then adding progressively more flexible ones. We redesigned the software around that process.
The current system divides the work rather than attempting complete automation. The software creates a base layer from calendars, rosters, vacations, and preset assignments. I add discretionary assignments based on that month’s needs, after which the software fills remaining openings while checking assignments against the rules. As I review and fine-tune the schedule, it tracks requirements and flags discrepancies.
This approach has reduced my monthly scheduling time from approximately 60 hours to 30. More importantly, it has changed where that time is spent. I can focus more on resolving conflicting requests, balancing call teams, reviewing schedules for reasonableness, and communicating with residents. We have also added call heat maps to visualize assignment distribution and a Google/Outlook-compatible calendar interface that residents can share with partners and family members.
The hardest scheduling problems remain human. What makes a schedule good or bad is partly subjective. Fairness may depend on assignments made months earlier, and one unexpected personal event can create cascading changes across an otherwise balanced schedule. While software can flag a conflict, deciding how to resolve it still requires the scheduler’s judgment.
A residency schedule is more than boxes on an Excel sheet. Each assignment represents a physician’s time, the staffing needed for patient care, and a portion of someone’s life outside the hospital. We began by trying to teach a software program how to build a schedule. What we ultimately built was a system that supports calendar and roster maintenance, consolidates information, and checks scheduling requirements, leaving the scheduler more time for decisions that still require a person.
(#2).
Building a Mock Oral Examiner: An LLM-Based Platform for ABA SOE Preparation
Joshua E. Insler, MD
Mentors: Marc A. Sherwin, MD,Viji Kurup, MD
Department of Anesthesiology, Yale School of Medicine, New Haven, CT
Preparation for the ABA Standardized Oral Examination (SOE) is often constrained by faculty availability, scheduling, and the limited number of practice opportunities most programs can offer their trainees. To address these limitations, we assembled a team of anesthesiology faculty, residents, and a biomedical engineer to develop an artificial intelligence (AI)-based mock SOE examiner using a large language model (LLM) and sought to document the process for other programs interested in doing the same.
We built our platform using OpenAI’s Chat Generative Pretrained Transformer (ChatGPT) ecosystem (GPT-4o, July 2025 version, OpenAI, San Francisco, CA), which allowed us to configure GPT-4o's behavior through system-level instructions and an uploaded knowledge base of publicly available materials published by the American Board of Anesthesiology (ABA), without any custom software development. Native voice-mode functionality allowed us to replicate the spoken format of the actual examination. Our goal was to create an accessible, low-cost tool that could supplement— not replace— faculty-led mock oral examinations by giving residents additional opportunities for deliberate practice.
Our most significant developmental challenge and, in turn, our biggest lesson, was shaping the LLM to behave like an examiner rather than a tutor. By default, GPT-4o had been optimized to be helpful and affirming, which is at odds with the neutral, probing demeanor expected of an SOE examiner. In early testing, the model consistently offered unsolicited encouragement and educational explanations mid-examination. Refining our system prompts to suppress these tendencies required substantial iteration and we found that instructions provided to the LLM specifying concrete, observable actions we expected from it produced significantly more reliable behavior. Once examiner persona and structure were stable, we developed fictional clinical cases modeled on SOE scenarios published on the ABA’s website, each incorporating several intraoperative complications – just as examinee’s could expect on test day. Working with faculty experienced in SOE preparation, we also developed behavioral anchors to guide the model's assessment across the ABA's five SOE scoring domains (since no formal scoring rubric is publicly available). At the conclusion of each session, the platform provided structured feedback that referenced specific moments from the examination, offering a level of individualized detail that can be challenging for faculty to consistently provide to trainees at the conclusion of each administered mock exam.
The completed platform requires no specialized infrastructure and is readily replicable by other programs with access to a ChatGPT Plus subscription. This work was presented at the 2025 Yale University Ideathon and has since been submitted as original research to the Journal of Medical Education and Curricular Development. We are currently planning a prospective evaluation of the platform's impact on resident confidence, exam-related anxiety, and first-time pass rates. Furthermore, we hope this serves as a useful and replicable starting point for other programs seeking to expand access to oral examination practice using tools already available to most residents and faculty.
(#3).
Physicians Should Have a Seat at the Algorithm, Too
Aesha Patel, MD
Mentor: Bryan Mahoney, MD
Mount Sinai Morningside and West Anesthesiology, Perioperative & Pain Medicine
I did not become interested in artificial intelligence because I thought medicine needed more technology. If anything, medicine already asks us to interact with too much of it. My interest came from the small frustrations of residency: feedback that never makes it into an evaluation, minutes lost searching for information we technically already have, and tools that seem designed around everything except the moment we actually need them. These are small inefficiencies but repeated every day; they shape how we learn and practice. As AI moves rapidly into medicine, much of the conversation has focused on what it might replace. I find a different question more compelling: What becomes possible when physicians themselves can build with it? Until recently, recognizing a problem and building software to address it were two very different skill sets. AI-assisted development has begun to narrow that gap. It has allowed me to move from saying, “Someone should make this better,” to asking, “Could I make this better?”
That question has led to two projects during my residency.
- “Rate the Residents” reimagines resident evaluations as a brief conversational interaction, making it easier for attendings to give meaningful feedback without another cumbersome form. Rather than asking the evaluator to conform to the evaluation form, could the evaluation conform to the way an attending already thinks and speaks? With the application, an attending simply selects the resident and completes the evaluation through a brief conversational interface, closer to telling a colleague how someone performed than filling out another institutional form. The evaluations can then be organized and reviewed through an administrative dashboard.
- “AnestheSync” grew from a different frustration: synthesizing the scattered and remarkable information needed to prepare for an anesthetic, from the patient's physiology and comorbidities to the operation, airway, positioning, medications, monitoring, ventilation, and contingency planning. I built it around the way anesthesiologists actually prepare. A resident can enter an upcoming procedure and build a case-specific anesthetic plan while accessing integrated anesthesia textbooks, ventilator presets, and regional anesthesia guidance with NYSORA ultrasound images, bringing resources that would normally require several separate searches into one place.
Building these tools has made AI feel both more powerful and more consequential. Once technology enters a clinical or educational workflow, questions of bias, privacy, overreliance, and accountability are no longer theoretical. Physicians should be part of that answer. That is why I believe physicians need more than a seat at the receiving end of this technology. We should have a seat at the algorithm, too.
We occupy an unusual position. We understand the clinical environment, but we also experience its inefficiencies from the inside. Residents, in particular, spend their days moving between the educational and clinical sides of medicine. We see where information gets lost, where workflows create unnecessary work, and where a seemingly minor change could make it easier to learn or care for a patient. AI will almost certainly become more deeply woven into medicine. I do not know exactly what that will look like, and I am skeptical of anyone who claims to. What excites me is the possibility that physicians do not have to be passive recipients of whatever technology arrives. We can help decide what gets built. And, increasingly, we can build some of it ourselves.
(#4).
Bringing Airway Education to the Undergraduate Curriculum in Colombia
Sebastian Amaya1,3 MD, Juan Sebastian Rivera Becerra2,3 MD
Mentor: Marc A. Sherwin, MD
Department of Anesthesiology, Yale School of Medicine, New Haven, CT
1. Yale School of Medicine, Department of Anesthesiology, New Haven, Connecticut
2. Detroit Medical Center/Wayne State University, Detroit, Michigan
3. Universidad El Bosque Colombian School of Medicine, Anesthesiology and Critical Care Interest Group, Bogotá, Colombia

In many countries, including Colombia, physicians may graduate from medical school and practice independently, without residency, and find themselves in settings where an anesthesiologist or airway specialist is not immediately available during an emergency, especially in a rural setting. Yet exposure to structured airway management during medical school can be limited. Recognizing this gap, two resident physicians and collaborators from Universidad El Bosque, Dr. Sebastian Amaya (Yale) & Dr. Juan Sebastian Rivera (Detroit Medical Center), developed a dedicated airway education and simulation workshop for medical students.
The initiative grew out of a national survey of Colombian medical students, done by Dr. Amaya and the Colombian Society of Anesthesiology and Resuscitation, that identified limited exposure to fundamental airway skills, including bag-mask ventilation, supraglottic airway placement, and endotracheal intubation. The idea was to create an opportunity for students to develop both foundational knowledge and hands-on experience with airway management.
The workshop is structured as a full day of education. The morning consists of focused lectures covering airway anatomy and assessment, indications for intubation, basic and advanced airway devices, peri-intubation medications, and difficult-airway management. In the afternoon, students rotate through hands-on stations practicing bag-mask ventilation, supraglottic airway placement, and endotracheal intubation before participating in simulated clinical scenarios involving airway emergencies and failed-airway rescue.
To expand the simulation component, Dr. Rivera developed Aetheris, a browser-based perioperative simulator built around a live physiology and pharmacology model. The platform allows learners to practice clinical decision-making in realistic cases, with patient physiology changing dynamically in response to their interventions. It currently includes 39 cases across multiple subspecialties and provides structured debriefing after each scenario.
By combining traditional teaching, hands-on airway skills, mannequin-based simulation, and virtual clinical scenarios, the program aims to give medical students repeated opportunities to recognize airway problems, initiate appropriate management, and understand when they have reached the limits of their capabilities.
Our goal is not to turn medical students into anesthesiologists. It is to ensure that physicians who may one day encounter an unanticipated airway emergency have had meaningful exposure to the principles and practical skills needed to respond safely. We hope this model can serve as a starting point for incorporating more structured airway education into undergraduate medical curricula in Colombia and other settings where physicians may practice without immediate access to airway specialists.





