Three Months, Two Colleges, One Honest Reckoning — What the EdMedAI Pilot Taught Me

I have built systems, written curricula, and delivered lectures across two countries. But there is a particular kind of education that only a real-world deployment can provide — the kind where your assumptions meet actual humans, in actual institutions, doing their actual jobs. That is what the three-month pilot gave me.

This is my honest account of what happened when EdMedAI went live at Kurnool Medical College in Kurnool and Pinnamaneni Siddhartha Institute of Medical Sciences in Guntur — both in Andhra Pradesh, both facing the same NMC compliance pressures around competency based medical education, and both with very different institutional personalities. I am writing this not to market the platform, but to think through what I observed — and to share the lessons that will shape everything we build next.

1. Why These Two Colleges

The choice of pilot institutions was deliberate. I wanted colleges that were not already digitally sophisticated — not early adopters who would validate a technology simply because they were predisposed to do so, but institutions representing the real mainstream of Indian medical education: dedicated faculty under significant administrative load, students navigating a demanding curriculum largely without personalised support, and principals who cared deeply about NMC compliance but had limited tools to achieve it systematically.

Together, they gave me something close to a genuine cross-section: one college with deep institutional roots and strong faculty culture, and one with management urgency and a clear compliance mandate. If EdMedAI could work for both, it could work for most of the 816 medical colleges the NMC regulates.

2. The First Week: Onboarding Was the Real Test

I have led digital health implementations in American hospitals. I have watched EMR rollouts, clinical decision support deployments, and analytics platform launches. I thought I knew what onboarding looked like. The first week of the pilot recalibrated me.

At Kurnool Medical College, the initial session involved faculty across multiple departments — surgery, medicine, pathology, community medicine. These were people who had been running their own paper-based CBME workflows for years. Their DOAP records were in physical registers. Their student logbooks were physical books with handwritten entries. Their attendance records were sheets maintained by department attendants.

What I saw in that first week was not resistance — it was something more nuanced. It was the very reasonable caution of people who had been handed digital solutions before that did not actually solve their problems. They were not hostile to EdMedAI. They were watchful.

At PSI Medical, the management had briefed faculty in advance, so there was a higher baseline of readiness — but a different challenge emerged. Faculty were willing, but the volume and specificity of what EdMedAI captures initially felt overwhelming. Logging a DOAP session requires indicating which student, which competency code, which stage (Demonstrate, Observe, Assist, or Perform), and the date and context. For faculty accustomed to signing a paper register, this level of granularity was unfamiliar.

By the end of week one, I had revised my plan. The training sessions had to do less explaining of features and more showing of value — specifically the question: "What does this system give you that you did not have before?" Once I framed it around that question, the conversations changed.

3. What Faculty Discovered in Week Two

The shift happened around the second week, and it happened in a way I did not predict.

At Kurnool Medical College, a Surgery faculty member — one of the more sceptical voices in the initial session — came back after three days and said something I wrote down immediately: "For the first time, I can see all my students' progress in one place. Not just who attended what session — which student has actually done what, and how many times."

What he was describing is something that sounds simple but is actually a profound operational shift: visibility. Faculty in Indian medical colleges are responsible for enormous student cohorts, and under a paper-based system, they have almost no real-time visibility into where any individual student stands. They cannot easily tell which of their 60 surgery students has completed the Perform level on wound closure versus which ones are still at Observe. EdMedAI makes that visible — not just for one competency but across the entire subject.

At PSI Medical, the insight that changed faculty behaviour was different: the AI content generator. A faculty member preparing lecture slides on abdominal trauma tried the AI Lecture Plan generator — sceptically, expecting something generic and unusable. What it generated was a detailed, structured lecture plan mapped specifically to the relevant NMC competency code, with slide-by-slide content, estimated duration per slide, and teaching points. She said the time from "blank page" to "usable draft" had dropped from several hours to under thirty minutes. She shared the output with two colleagues. Those two colleagues tried the tool the next day.

4. What the Students Showed Me

I had modelled student behaviour on a hypothesis built from years of working with medical students: that they would engage with digital tools enthusiastically but superficially — using the platform when required, not beyond it. The pilot contradicted this at both colleges.

At Kurnool Medical College, students adopted the digital logbook faster than any other user group. Within the first two weeks, students were not just logging required entries — they were using the self-assessment features, adding learning points to their procedure logs, and in several cases initiating their own DOAP session documentation ahead of faculty reminders. The average detail in student logbook entries was higher than I had designed for. Students were writing more than the minimum required fields — treating the logbook as a genuine record of their learning, not a compliance form.

At PSI Medical, the engagement came through a different channel: the AI Tutor and the MCQ quiz system. Students were using the AI Tutor to ask questions about clinical concepts at hours — late evenings, early mornings — when no faculty member would be available. The spaced repetition quiz system was being used voluntarily, not just when assigned. A small group of final-year students told me directly that having a quiz system that "knew which topics they were weak on" was something they had never experienced before.

5. The AETCOM Module — The Surprise of the Pilot

I want to talk about AETCOM specifically, because it produced the finding that surprised me most. AETCOM — the Attitude, Ethics, and Communication module of the NMC curriculum — is genuinely difficult to digitise. It is about professional identity formation, ethical reasoning, and communication skills. It requires reflection, not just documentation.

At Kurnool Medical College, several students submitted AETCOM portfolio entries that went well beyond the minimum required content. One entry, written by a first-year student after an early clinical exposure session, reflected on the dissonance between the clinical efficiency they had observed in a busy outpatient ward and the patient communication principles they were being taught. It was thoughtful, specific, and showed genuine professional reflection — the kind of thinking AETCOM was designed to develop.

I showed this entry to the faculty facilitator who had reviewed and approved it. Her response: "Before this system, I had no way to see what students were actually thinking about their clinical experiences. I could see who attended the session. I could not see what they took from it."

That observation sits at the heart of what competency based medical education is supposed to do — shift the focus from attendance to learning. And it is what the digital AETCOM module, used with intent, actually delivers.

6. The Moments That Were Hard

I would be doing a disservice to anyone reading this if I described the pilot as straightforwardly smooth. It was not.

The attendance system required more configuration than expected. Getting geofence boundaries right for clinical posting locations — which change by semester and by department rotation — was genuinely complex. At one point early in the pilot, a cohort of students could not mark attendance for a clinical posting because the geofence had been set to the main campus rather than the affiliated hospital. Faculty had to manually update attendance for that session. We fixed the geofencing configuration workflow within the same week — but it underlined that geography in Indian medical education is complicated. Colleges operate across multiple campuses, multiple hospital affiliations, and clinical posting schedules that change frequently.

The parallel-system problem. In the first four weeks, several faculty members at both colleges flagged that entering DOAP session records was adding to their administrative burden — because their departments had not yet retired the parallel paper system. They were running both paper and digital in parallel, which doubled the documentation work. It required me to work directly with HODs at both colleges to establish a clear changeover date — a point after which paper registers would no longer be the system of record. Once that commitment was made, the parallel-running stopped and the burden dropped immediately.

Connectivity. Several clinical posting locations at affiliated hospitals had unreliable internet access, which affected real-time attendance marking and logbook uploads. We adapted quickly — adding offline-tolerant patterns to the most critical workflows. But it was a reminder that EdMedAI is being built for institutions that operate in a country with significant infrastructure variability. A platform that requires constant connectivity is not a platform that will work for all 816 NMC colleges.

7. What the Data Showed by Month Three

By the end of the three-month pilot, the usage patterns across both colleges had settled into something consistent enough to draw real observations from.

Faculty engagement had shifted from initial scepticism to routine use. The most-used faculty features, in order, were: logbook review and sign-off, DOAP session logging, the AI content generator for case studies and lecture plans, and the attendance management dashboard.

Student engagement was broadest in the quiz and logbook modules. The AI Tutor was used most heavily by second and third-year students. Clinical simulation usage was lower than hoped — not because of disinterest, but because integrating simulation exercises into a busy clinical timetable requires coordination at the HOD level that had not yet been established at either college.

HOD adoption was slower than faculty adoption, but the data was compelling when they engaged. One HOD at PSI Medical described it this way: "Before this, I was writing my NMC inspection report from memory and from whatever records faculty could assemble. Now I can generate the compliance summary in twenty minutes. The inspector can ask me anything and I can show it immediately."

8. What Each College Taught Me That the Other Did Not

Kurnool Medical College taught me the power of institutional culture. When the Surgery department adopted EdMedAI seriously in week two, other departments watched and followed. Peer credibility inside an established institution is a more powerful adoption driver than external advocacy. This is something I will build directly into future rollout plans: identifying and engaging with the most respected voices in an institution first, before broad deployment.

Pinnamaneni Siddhartha Institute taught me the importance of management alignment. When the leadership team made it unambiguous that EdMedAI was the institutional direction — not a pilot to be evaluated indefinitely but a decision that had been made — faculty adoption accelerated in a qualitatively different way. The ambiguity that slows adoption ("is this going to stick?") was removed early. Faculty invested in learning the platform properly because they knew they were investing in something permanent. Clear management commitment is probably the single biggest lever available to any institution deploying a new educational technology.

9. Revisiting the Vision — What Has Changed

When I designed EdMedAI, I was working from a model of the problem built from the outside: from visiting colleges, speaking with faculty, reviewing NMC inspection reports, and drawing on my experience of what a functioning competency based medical education system looks like in American graduate medical education. The pilot gave me the inside view.

I had underestimated how much of the implementation challenge is social rather than technical. The platform itself — its architecture, its NMC alignment, its AI capabilities — worked substantially as designed. What required the most active management was not the technology but the human transition: helping faculty see that the system was not surveillance but support; helping students understand that logging entries was building a record that would serve them; helping HODs recognise that real-time visibility was an asset, not an exposure.

I had also underestimated the appetite for AI-powered learning at the student level. The quiz system and AI Tutor usage patterns across both colleges suggest that Indian medical students are significantly more ready for personalised AI-assisted learning than the broader education system has been willing to offer them.

10. What I Am Building Next — And Why the Pilot Determined It

• 1
  Offline Resilience Connectivity challenges in affiliated hospitals are real and persistent. The platform must function reliably in low-connectivity environments, syncing when connectivity is restored, without data loss or workflow interruption.
• 2
  Multi-College University Management Both pilot colleges are part of university affiliations. The conversations I had with management were not just about their own college — they were about extending the platform across all affiliated colleges. University-level consolidated dashboards and cross-college compliance reports are now a near-term build.
• 3
  Faculty Champion Programme The pilot demonstrated that the most effective onboarding happens peer-to-peer, not through formal training. A structured faculty champion model — identifying two or three early adopters per department and equipping them to bring colleagues along — is now the standard rollout design.
• 4
  Simulation Scheduling Integration Low simulation usage during the pilot was almost entirely a timetabling problem. Integrating simulation exercises directly into the timetable — with HOD-level scheduling tools — will resolve this.

11. To the Faculty and Students of Kurnool Medical College and PSI Medical

To the faculty at both colleges who gave the platform a genuine try — who entered their DOAP records carefully, who reviewed and signed student logbooks, who used the AI tools to prepare teaching content, who flagged problems directly rather than working around them — thank you. You made this pilot honest. The frustrations you expressed, the workarounds you needed, and the features you didn't use are as important to me as the features you did.

To the students who filled in their logbooks thoughtfully, who used the AI Tutor at 11 pm before an assessment, who wrote AETCOM reflections that were genuinely reflective rather than formulaic — you are the reason this platform exists. Every feature decision I make is, ultimately, a decision about what will make your training better.

And to the principals and HODs who took the institutional risk of being first — who committed to a platform that had not yet been proved in production, in real Indian medical colleges, under real compliance pressure — you gave me the most valuable thing a founder of an early-stage product can receive: the chance to find out what is actually true, rather than what I assumed.

12. Three Months In — What I Believe Now

Twenty-five years of working in American medical education gave me a framework. Three months of real deployment in Indian medical colleges gave me evidence. The two things are now integrated in my thinking in a way they could not have been before.

I believe — more strongly than before the pilot — that the quality gap in Indian medical education is not a talent gap. It is a systems gap. The talent is there. The students are capable. The faculty are dedicated. What has been missing is the infrastructure of accountability and support: the real-time visibility, the structured documentation, the personalised learning tools, the feedback loops that close the distance between what a student has done and what a faculty member knows about it.

I also believe that Indian medical colleges are more ready for this change than the broader narrative about technology adoption in traditional institutions would suggest. The rate of genuine, sustained adoption at both Kurnool Medical College and PSI Medical — at the faculty level, at the student level, at the HOD level — was faster and deeper than my most optimistic projection.

The path from two colleges to eight hundred and sixteen is long. But the pilot answered the foundational question — the one that any honest founder must keep asking: does this actually work, for real people, doing real work, in the real conditions they face?

The answer, from three months and two colleges in Andhra Pradesh, is yes. It works. And we are just getting started.