Physicians Belong in AI: A Founder’s Path from Academia to Impact with Dr. Shakira Grant

On this episode of Startup Physicians, I interview Dr. Shakira Grant, who shares her remarkable journey from practicing medicine to influencing healthcare policy and AI development. We talk about the importance of health equity and the role physicians must play in shaping AI tools that serve all populations effectively. Shakira reflects on her time in Congress, the challenges of integrating AI in clinical care, and why diverse voices are essential in the development process. She encourages physicians to embrace innovation and step into roles where we can use our expertise to drive meaningful change in healthcare outcomes. 

Episode Highlights:

[00:00] - Introduction to Dr. Shakira Grant's Journey
[02:46] - Transitioning from Medicine to Policy
[05:39] - The Intersection of AI and Health Equity
[08:40] - Challenges and Opportunities in AI Development
[11:28] - The Role of Physicians in AI Innovation
[14:13] - Founding CROSS: Bridging Healthcare and AI
[17:05] - Practical Involvement of Physicians in AI Projects
[19:58] - The Importance of Diverse Voices in AI Development
[23:05] - Final Thoughts and Encouragement for Innovators

Transcript

Alison Curfman (00:00.821)
Hi everyone. Welcome back to Startup Physicians. This is your host, Dr. Alison Curfman, and I am joined today by Dr. Shakira Grant, who has a really incredible story to share. And I knew she was going to be a good podcast guest because I was introduced to her not once, but twice. Somebody wanted me to know her. so Shakira, thank you so much for joining me today.

Shakira J. Grant (00:25.154)
Thanks so much for having me. I'm excited for our conversation.

Alison Curfman (00:29.129)
Yes. So in our chat about your background and your journey and all of these cool things that you've done, I'm so excited for people to hear about how you got to where you are now in the work you're doing today. think those of you who stay on to listen to this whole episode, Shakira has some really phenomenal vision of how to help empower doctors.

to make an impact in developing AI healthcare technology. so we'll get to that. But before we get there, can you tell me a little bit about your medical career and how things started and when things shifted for you?

Shakira J. Grant (01:12.11)
Sure. So I've been in medicine now for almost 20 years and over that course I have largely provided direct patient care starting out from my training in Barbados and where I practice medicine. At the heart I will say it has always been about serving people, ensuring that communities that normally do not have a voice or sit at the margins that they are represented and that has always been

my purpose and that has guided my career. And so I've always been the type of physician who never was quite satisfied with status school. This meant that I was oftentimes one only different in the room, but never afraid to step outside of the box and carve a new path. And that led me to a career as a geriatric oncologist. And while I was at the University of Washington,

And then ultimately I went the academic route initially and started my research lab and got really great funding, supported over 10 reportees and trainees and was really by all means successful. But at the core, there was always this calling that I wanted to do more outside of the clinic. I knew what I was doing in the research could affect maybe 50, maybe a hundred people.

But I had this calling to really think bigger on the population level. And that is what led me to make some serious shifts in my career.

Alison Curfman (02:47.521)
That's incredible. I really very deeply to those words. So tell me what was the pathway you took after that or what opportunities did you look for?

Shakira J. Grant (02:56.59)
Yeah. So while I was still at UNC Chapel Hill as faculty, I really continued to work on health care access and disparities. And I remember so clearly I had spent 2022 for the entire year actually writing a grant, an NIH grant for one year. It was called a Beeson. And so this is a very prestigious K award. And

Lo and behold, I got a great score, which meant this grant was gonna be funded five years, $1.2 million. And I decided to walk away. I left it on the table. Yes, so ha, yeah.

Alison Curfman (03:41.825)
Sometimes a career pivot, you're leaving something behind, but there's something much more ahead. So what prompted you to do that?

Shakira J. Grant (03:51.277)
Yeah, so it was the opportunity of fellowship through the American Society of Hematology to serve in Congress initially for one year. And at the time I had no policy ideas, I had never been exposed to policy. And I just saw this as an opportunity to say that thing that has been nagging at me to say, I don't want to necessarily only think about who I can implant within the clinic walls or within the walls of one institution.

wanted to go bigger. I wanted to impact on the population level and I took the leap. I left that grant that was likely to be funded by the NIA on the table and I walked away from academics and I walked right into Congress. And that led me to the Committee on Ways and Means, which for viewers, if they don't know, this is one of the most powerful committees in Congress. And this is where everything related to taxation originates in the U.S. House of Representatives.

Alison Curfman (04:50.647)
That's incredible and a very unexpected turn. it reminds me of my journey in some ways too, that it's like, what, what am I doing? get what I, this was not on my roadmap. Like five years ago, I would have never thought that anything like this would be on my pathway. And, and like many people I've talked to really emphasizing that when you enter a new realm like this, there's so many new.

skills to develop and language to understand and you're just like a fish out of water and really just absorbing so much. But what did you take from that period in your career and what sort of impact did you feel like you were able to have?

Shakira J. Grant (05:32.963)
Yeah, so in Congress, one of the big things that I learned from day one was how to lead when you are not in charge. That was the first thing, because I was accustomed to walking in spaces where as a physician, you walk into the room and everyone is looking to you for the expertise. But I was now the person who was the most junior on the team and not necessarily.

having any policy experience, so everyone else was so much more well-qualified than I was. And so that experience, it really brought me to a place of humility, which I really enjoyed being in that space because I understood what it took to work day in and day out on behalf of the American people, especially when many of the congressional staffers are doing this work behind the scenes. And so you may never know their names.

but they are like the driving forces behind all of this. The other I would say, it taught me a lot about working collaboratively and how to communicate when people do not have the same background or training as myself and how to communicate and break down science, especially in a way that people who are writing policies and dealing with these issues on a large national scale could actually understand. So there were so many good, good.

takeaways, but I really learned a ton of leadership skills, communication skills, and really how to work in cross-functional teams.

Alison Curfman (07:04.491)
That's really important. it sounds like along that journey, you learned a lot about AI as well and how it was impacting healthcare.

Shakira J. Grant (07:16.972)
Yeah, yeah. So one of the things with being on Ways and Means, their jurisdiction really lies for Medicare beneficiaries. And so there are about 66 million Medicare beneficiaries in the United States. And so that's a large population. These are largely older adults or people living with disabilities who are under the age of 65. And part of the work that I was able to do there was really to bring that intersectionality between

health equity, which was my background, and bringing that into this space to think about artificial intelligence. And so what that looked like is if you think about age as one of those marginalizing factors, so older adults especially, this led us to really think about policies in this area. What are some of the policies as it relates to AI that can potentially impact Medicare beneficiaries? And that is where I really got the first exposures to

artificial intelligence and digital health from this population lens or policy lens. And that just led me to really want to do more in this space because while I saw a lot of the potential, I also saw the pitfalls and I saw again that there was an opportunity for populations to be left behind.

Alison Curfman (08:34.455)
Yeah, oh, absolutely. I mean, think as much as we're all excited about the different things that AI could do in our lives, I think a lot of us have a lot of fear as well. Like, what is this going to do to our field? And I don't so much know doctors that are like, oh, no, will I be out of a job? It's more like, will this affect quality and safety? Like, is this going to worsen health equity issues? I actually was recently asked to review a manuscript for some

academic publication. I don't even know who the authors were because it was blinded, but it was about the inherent bias that was present in a lot of the AI models and the way it was recommending things differently, like healthcare things differently for different populations inappropriately. And so I think a lot of us have a bit of fear, healthy fear of what AI is going to do to medicine. And I'm sure that crossed your mind as well, but that prompted you to

Shakira J. Grant (09:19.342)
Mm-hmm.

Alison Curfman (09:34.207)
not just cower in fear, to actually try and create something for that.

Shakira J. Grant (09:36.355)
Yeah, exactly. So I think you can take two approaches in this AI driven world that we are seeing. And I think one of the biggest things that I see that I feel a bit of skepticism, fear, anxiety around is the speed at which tools are being developed and actually translated right into healthcare settings. This is a speed that we haven't seen before when we bring new drugs or new therapies to market where we often expect to see

rigorous data behind them before actually wanting to test them out. But now it's almost like we are pushing so much for innovation that we're not slowing down enough to think about these biases. And it's funny you mentioned this because I was just at a conference and gave a talk this weekend and part of it was really talking about can AI be the thing that closes disparities or further widens them. And the example I gave was about using

tools to detect skin cancer that underperform in people with darker skin tones because the model was not trained on people with darker skin tones. And so it misses skin cancer. So imagine just using that tool, which there are these tools out there. This isn't hypothetical. And missing skin cancer, for instance, in Black populations who already have worse outcomes if they're diagnosed with melanoma. So yeah.

Alison Curfman (10:49.889)
Hmm.

Alison Curfman (11:04.586)
Right, but I'm gonna actually challenge you that it's not just like, the AI tool is not as good at that. I think our entire medical system, this is a known health equity issue that like in dermatology, like we are not as good at diagnosing things in brown and black skin. We are not as well-trained. Even like I've seen studies about like the number of like texts and the...

know, visual representation of all these different pathologies, like that they're so skewed towards showing white skin that like, it's almost like it's exacerbating an underlying issue that exists in the way we practice medicine today. But since it's a technology tool, the question is, how can we actually, like it would actually take quite a bit to really bring every doctor in the US and every clinician up to par to really have

Shakira J. Grant (11:33.666)
question.

Alison Curfman (12:00.791)
the same level of diagnostic skills in both types of skin or all types of skin. But I guess the question is now, can we make this better? Can we actually reduce health inequities by training models better or by being aware of it? But it needs to be very focused on

Shakira J. Grant (12:17.613)
Yeah, right.

Shakira J. Grant (12:25.219)
Yes.

Alison Curfman (12:26.411)
this being a problem and acknowledging that health equity needs to be a number one goal of what we're doing with AI.

Shakira J. Grant (12:35.938)
Yeah, and that is challenging, right? It's challenging to get people to even use the word health equity bias currently in this current political climate, right? Because those words have been politicized and made to seem as if this work doesn't matter. But here's why, to your point and to your question, you are absolutely right that, yes, there is bias in the tool, but there's also different areas where the bias is going to come in.

So the model is trained on historical data that is already encoded with biases. So that is the way that you see the model doesn't understand when it needs to detect skin cancers, for instance, in darker skin tones. The other is the human bias element. that is the real issue here in terms of how historically, especially when you look at things like

maternal mortality, for instance, and other diseases even within hematology like sickle cell disease and how those patients experience bias, especially within the healthcare system. When you add that on top of tools that are scaling these inequities now, so before it may have impacted maybe a few hundred maybe, but now you have models that are being widely disseminated.

So these inequities now are reaching scale. They're being compounded by human biases, as well as other healthcare biases that exist. And so when I think about AI's promise for healthcare and reaching health equity, I think it needs to be intentionally paired with a mirror that says, this is the reality of our healthcare system, and these are the biases, and these are the intentional steps that we need to take.

to make sure that we are correcting those biases beyond just looking for technology developers to correct the data.

Alison Curfman (14:29.025)
Yeah, for sure. And I think that when I work with doctors on how they can develop skill sets to work with startups, not just AI startups, but like any startups, I mean, one of the things that we struggle with as physicians is sometimes being too rigid of thinkers and having trouble thinking about new and different ways we can do things. But I encourage people to have very divergent thinking and really be willing to brainstorm and push their boundaries.

to always stay true to quality and safety and health equity as like our core foundations of what we bring as physicians to industry is a deep understanding of like, is this safe? Is this high quality or not? And is this going to make things more equitable or less equitable? And I think there's such an important space for physicians' voices in this area.

Shakira J. Grant (15:27.79)
Yeah, no, I agree. And I think when people think about what role can a physician's play in terms of this new and evolving AI role, I say be a part of this from every single stage of the AI development life cycle. So all the way from concept. So somebody has a great idea. Clinicians' voices are essential. And not only just

having the same clinicians who you're accustomed to, but diverse representation. People who have a voice and a foot in the communities that can bring the real issues that communities are struggling with so that they can be elevated and be the focus of some of these tools that are being developed. And so when you think about where you can plug in as a clinician, I say every single step a clinician's input and voice is definitely needed.

because you can shape the ideas, you can shed light on these inequities, you can help at the data stage and say, encourage people to look at the types of data they're using to develop and build their models. And then you can help with the implementation. So I say bias is one aspect, but when you think about equity, we need to also step back and say, who is benefiting from all of this innovation? Because if you even look across the US,

their health care systems, especially those that serve marginalized groups, those that primarily see, for instance, Medicaid populations, they're not able to afford these new and exciting health AI innovations. So you're going to see these tools concentrated in very highly resourced settings, and the marginalized communities served by the federally qualified health centers, for instance, or the community clinics, they're going to be left behind. So we need to be thinking about

How do we close these gaps at every single stage in the AI development lifecycle?

Alison Curfman (17:26.391)
I think this is a great segue into discussing what you're working on now, which is on the one hand, I know you can speak to being a founder now and some of the doctors out there who might be listening that maybe have an idea or an inclination or a vision and kind of speaking to what it took for you to actually found your own organization. But then also going into like what sorts of opportunities you're trying to create for physicians to contribute.

to high quality AI in healthcare.

Shakira J. Grant (17:58.467)
Yeah, yeah. So I would say the first step of being a founder and having an idea is to do the thing. Just do the thing. You will come up, believe you me, with so many different reasons why you should not start that thing, why you shouldn't pursue it. The only way you can tell if your idea will actually work is by doing it. And believe me, you will fail. And you fail, you learn, and you iterate. It's a

process that you have to go through as a founder and I can't stress that enough and I think one of the things that I've learned in this journey as I you know talk to people is people are just so scared of failure that they remain stuck and they say I had an idea but you know was too scared to take the lead so I just continued in my path the one that I know that is well walked and well frauden but for me I was always

Let me just think about how I can build a new path and make a way for other people. So that led me to CROSS. And when I think about CROSS, global research and strategy, it's a mouthful, but I abbreviate it as CROSS. And it really is meant to be the intersection, the intersection of all that I've been building, all the experiences and clinic research and policy coming together to really shape equitable AI development and implementation across populations.

And what CROSS does is we work directly with people who are building health AI tools. So this means from early stage startups all the way to big tech. And some of the roles could be from providing advisory services to just focus on your subject matter expertise, all the way to more the technical aspects, such as being an AI tutor to help guide the types of outputs that the model is actually giving.

And so that is kind of where CROSS is and sitting at this intersection and really being the bridge between saying, let's bring the clinic to the code, right? So the technologists are developing these tools in silo. Let's break down the silos and let clinician voices be represented within those silos.

Alison Curfman (20:09.621)
That's incredible. And I mean, I think that it's pretty rare for physicians to have experience working with development teams. But I think that there's a few things that are interesting about AI. So one, it's like the first technology where the native language is English. So like, you can actually interact with these models. Like if a dev team wants you to like give feedback to a model, you can actually interact with the model. It's not just bunch of ones and zeros.

But you described to me some examples of like what exactly a physician could do with an early AI company or someone who's trying to develop a model and making it better quality. Like, could you go into that a little bit about what that actually means or what it actually looks like?

Shakira J. Grant (20:52.59)
Yeah.

Shakira J. Grant (20:56.226)
Yeah, so I would say the first is to think about as a clinician, you already have a ton of transferable skills. And so that means that you know the pain points within the healthcare system. So you know what are some of the major issues and barriers that patients are facing or you from an operational standpoint are facing that you can actually feed that back to the developer's team.

You know the actual end result also in terms of how the healthcare system works. And so when teams are developing or proposing ideas, that's a critical juncture that you could say, you know what, this may work well in the lab, but to actually operationalize this, this is not possible. So that really falls under the bucket of clinical subject matter expertise, where you are serving as essentially the domain expert and helping to bridge knowledge gaps that may exist.

with the AI research labs or the other developer teams. The other aspect I would say is important is adversarial testing. And so adversarial testing or red teaming is another term you may hear, is basically trying to trick the model or take on different personas for a model that could be potentially used by bad actors when the model is released. So while we all have great intentions,

we recognize that models can be used for harm. And so a role a clinician could take is basically playing different roles and taking on that psychology of ways in which this could be potentially used in a malicious way and helping teams to actually realize that and actually correct those problems. And then the final one I would say is domain expert evaluation. So that is basically where you are being an AI tutor.

And so for instance, if the model is supposed to evaluate a patient with chest pain and it could be a heart attack, you want the model to actually recommend that this is potentially a heart attack or myocardial infarction. You don't want it to say that's toe pain, rear of shore and sent home. So that is some examples where clinicians could really get involved in the AI evaluation process and develop.

Alison Curfman (23:09.909)
I can't help but like laugh when you're talking about like the adversarial training. I'm like, would they pay me to fight with a bot? Like, or it makes me think of like when somebody pays a hacker to like, see if you can hack into our systems, you know, like, but, but it's just so funny. It's a totally different world, but yeah, you want to make sure that, these tools are trained on safety and you're coming at it from both angles. Like, are they recommending the right thing? And are they, are we preventing it from

Shakira J. Grant (23:17.907)
Red teaming.

Shakira J. Grant (23:23.894)
Yeah.

Alison Curfman (23:38.027)
doing the wrong thing no matter what comes at it. But yes, it's such a bizarre future we're going to be living in. But I I think of like when I was building a digital health company and we were going live, I remember thinking like always, constantly, what's the worst thing that could happen? What's the worst thing? Out of all of the things that we're doing, what could actually harm someone? Like how can we make sure that we never get into situation where someone is harmed by

Shakira J. Grant (23:39.234)
Yes.

Alison Curfman (24:08.139)
the innovative way that we're delivering care. And I think it's actually not hard for doctors to think about like, what's the worst thing that could happen? We're actually, I'm an ER doctor. I think about that all the time. Like that's every patient I see that I think is somebody that like is, you know, not that sick or it's pretty reassuring. always run, we run through a differential. Like what's the worst thing it could be and how am I convincing myself that

Shakira J. Grant (24:10.358)
Exactly.

Alison Curfman (24:36.255)
that is not what it is. What evidence do I have that this is, you know, reflux and not a myocardial infarction? And can we use that knowledge and share that knowledge with models as well to make sure that the tools that we're creating for healthcare are high quality? So your organization is crossglobalresearch.com.

And I now I very much understand why I was introduced to you multiple times and people were like, you need to know Shakira. Because I think what you're creating is so amazing because whether we like it or not AI is going to change healthcare. I believe that physicians need to step up. need to get involved. We need to stay informed. We need to use the knowledge that we already have to help inform the technology of the future.

Shakira J. Grant (25:09.038)
you

Alison Curfman (25:28.999)
And I'm so excited to be able to share with people what you're creating.

Shakira J. Grant (25:34.253)
Yeah, yeah, I agree. I'm with you that AI is here. I know there's a lot of fear and a lot of skepticism, especially people, maybe not so much in the clinicians, physicians space about job replacement. But the reality is if we do not have a voice and have a role to play as early as possible in developing these tools, then I ask myself, isn't that going against all that we

do and say that we are here to do as we are serving patients and there to ensure that their is paramount and especially when it comes to the equity angle. And so if you think about equity, if you think about safety and wanting to just do good by patients, then you can't just say, I'm afraid of AI. I don't want to be a part of this. This is where we need your voices because another set that we talked about a little bit briefly is the reality is there's

billions of dollars. Last year alone in 2024, VCs invested $11 billion in the health AI space. But guess what? 30%, only 30 % of AI tools actually get beyond the proof of concept stage. So 70 % are failing. Why? Because they're clinically poor workflow fit. They don't have the safety or the governance issues or guardrails sufficient to actually get implemented.

And a lot of this is being driven by not having clinicians or physicians in the loop early because to help guide those critical decision points that make that go no-go decision a lot earlier before you get down the pipeline of spending millions of dollars to develop a product that doesn't work on the market. So it's critical. Our voices are critical.

Alison Curfman (27:23.261)
That is such a powerful statistic. I think that it's just so telling that you who have spent so much time doing discovery with so many people in tech and industry and AI are hearing across the board, like, yes, we need physicians, we need clinicians, we need these voices. We don't want to be in the 70 % that never gets to market. And so there's a demand. sometimes doctors have trouble believing that they're

their skills could be useful or there could be a demand for their knowledge. So I love that you're being a connector for people and helping pave that road for other clinicians to make better technology of the future. So I would highly encourage anyone that wants to learn more. Shakira Grant on LinkedIn and then...

your website, cross global research.com I know has a contact form. So would that be a good way for doctors to reach out to you if they're interested in what you're doing?

Shakira J. Grant (28:25.1)
Yeah, yeah, yeah, please reach out. It's actually Shakira J Grant on LinkedIn. There's other Shakira Grants on LinkedIn, believe it or not. But yeah, yeah.

Alison Curfman (28:34.739)
Okay, yeah, we'll make sure to link that in the show notes as well, both of those. So I so admire all the work that you've done, the journey you've taken, this bizarre pathway from academics to Congress to starting an AI company. It's really incredible. Do you have any last thoughts to share?

Shakira J. Grant (28:46.655)
Yeah.

Shakira J. Grant (28:53.74)
Yeah, you know, I would say to anyone who is out there and tinkering with an idea that maybe you have this feeling deep down inside that you want to start the business and you're just not sure how to begin. Just just start putting your ideas out there. Just do the thing and just see what happens and learn from it. And because you feel

That doesn't mean you have to stop. It just means maybe you need to course correct. Maybe the idea you had is not the right one for the market at this time. And you just keep iterating and you keep going because you will find your path. But I think the biggest disservice we can do as physicians is to think that we went to medical school and the only thing that we can contribute is patient care. And I think patient care is valuable, but we also have to recognize that there are other lanes that we are.

quite capable of driving in and don't be afraid. Just do the thing.

Alison Curfman (29:51.393)
Thank you so much, very inspiring. And thank you for joining me today.

Shakira J. Grant (29:55.575)
Sure, thanks so much for having me. I've loved this conversation.

Alison Curfman (29:59.499)
All right, thanks.