Camille Morhardt 00:01
Hi. I’m Camille Morhardt, host of InTechnology podcast and very much looking forward to a conversation on silicon photonics and application for it in personalized medicine. This is an Intel Capital series podcast, so I’ll be co-hosting with Srini Ananth who is Managing Director at Intel Capital and focuses on deep tech, cloud infrastructure and AI applications investments. He’s led investments in all of these areas, and he also serves as a director or observer on the board of Astera Labs, IRE Labs, Common Sense Machines, Eleon, Inworld AI, VeriSIM Life and SiPhox.
Welcome to the podcast, Srini, and please introduce our guest for the day.
Srini Ananth 01:09
Thanks, Camille, and it is great to be here. Today, we have a great guest, Mike. Mike Dobrovsky is someone I met almost four years ago, actually, right after COVID hit and we were all in lockdown, and I happened to join a YC Demo Day, and that’s when I met Mike and his co-founder, Diedrik (Vermeulen), for the first time.
But Mike has a very eclectic background, as I would put it. He has an academic background in material science, and he’s spent time doing research at both Technion and MIT on biochemistry and photonics applications. He’s founded companies in energy tech, right out of college. He’s worked on a nonprofit that used a very innovative silicon photonics concept for cryptocurrency mining. And so, there is the energy efficiency angle right there. And currently he is the CPO and co-founder of SiPhox, an amazing company that is using silicon photonics and the economies of scale that comes with semiconductors in general to affect personalized medicine, personalized diagnostics. And great to have you here and looking forward to some amazing discussions here.
Michael Dubrovsky 02:09
Thanks, Srini. I’m excited to have this conversation.
Camille Morhardt 02:11
As Srini just mentioned, SiPhox is utilizing silicon photonics in the devices that it makes. And I know this technology is applied in a variety of different industries. Before we get into how you’re using silicon photonics, could you describe for us what it is and how it works?
Michael Dubrovsky 02:27
So, Silicon Photonics is really just, it’s like, transistors used to be these vacuum bulbs, right? So, it’s in terms of optics at some point it was all like, bulk lenses and things like that. And eventually, a lot of things went to fiber optics, so confining light in glass or other fibers with a core–and there’s an interior part and a shell, and you can find the light based on the refractive index difference. So, the only difference with silicon photonics is that it’s all being done in silicon structures or other chip-scale structures. And the really big difference there is that, in optics, alignments are very difficult, and any small variability can cause a lot of losses. It’s not like electricity, where you can take two wires and, like, touch them to each other in any direction, and you’re going to get electrical current flowing pretty efficiently; with optics, very small alignment–if you’ve ever seen a picture of like, you know, it’ll be an article like, you know, “Beijing University Builds an Optical Quantum Computer” or something, and it’s like, this optical table or seven optical tables, which are all with lenses and everything, and they’re aligned perfectly, you know, you look at that and you think like, “Oh, my God, these grad students probably spent like, three years aligning all of these lenses, right?” And so, the great thing about silicon photonics is that you get all of those things for free from lithography. Because the way chips are made, you have extremely high accuracy of the sizes of things and the alignments of things relative to each other.
If you’re able to fabricate successfully wave guides and other structures like basically fiber optic cables, but made of silicon on the chip, you can start making all of these components in a circuit that’s already pre-aligned, already pre-connected, so you don’t have to worry about all these the connections and the alignments and all these things. And also the precision of the components, normally, is very expensive, whereas in the case of a chip, you get the precision almost for free from the lithography equipment. So that’s kind of the promise of silicon photonics. And I think it’s really been very successful at that for telecom, which is probably the, you know, the biggest market for precision optics and lasers and so on—a t least at low power.
Camille Morhardt 04:28
But at SiPhox, you have a very different use-case, right? You’re offering at-home test kits where people can test biomarkers in their blood–like a better sense of their cardiovascular health, or nutrition or inflammation, thyroid, hormones, etc. And that puts you in both the medical and consumer spaces, which I think is interesting compared to other use cases for silicon photonics. And before we dive deeper into that, can you explain more about you using silicon photonics and the testing you’re doing?
Michael Dubrovsky 04:59
In the case of silicon photonics, it’s a resonant optical structure, so it’s something that absorbs a certain wavelength of light. In academia, people will attach proteins to it and make a sensor out of it. So, the one that’s been successful historically is called surface plasmon resonance. So, what that is a glass slide with a very thin sheet of gold on top, and they shine light underneath, and the reflection of that light, the angle of reflection changes when proteins attach to the surface. So, if you have proteins there and a liquid and something is attaching right at the surface, you’re going to get a change in reflection.
That’s used in pharma. So what they’ll do is they’ll, for example, if you look at like an antibody drug, you inject the person with a protein, and it binds something in their body and blocks it, or activates it, or whatever. A good example is like Ozempic. So Ozempic is a peptide. It’s like a short protein that binds something in your, like, a receptor in your body that it stops hunger. So, the way that they actually screen these things is they will attach your receptor—so they’ll take like the human receptor–and they’ll put it on that glass slide chemically, and then they’ll flow different peptides over it, and when they bind, you can see that in the reflection of the light. Instead of using a large glass slide, it’s taking a tiny component on a silicon chip, but doing the same thing–you attach proteins to it, and when those proteins recognize another protein, you see a shift in the resonant frequency.
Camille Morhardt 06:18
Thanks for explaining the science, Michael. And can you now talk more about SiPhox’s tests and devices?
Michael Dubrovsky 06:25
Basically, we realized that with the progress that’s been made in Silicon Photonics, it’s possible to commercialize it for the home. The overall idea is just to take some of the standard blood tests–the protein and hormone blood tests that are available now, from LabCorp and Quest–and make them extremely easy to do at home. So, it’s not just making a chip that does that, but all the readout equipment, like the laser, all the optics that connect to it, and everything, all of that can be miniaturized into a consumer electronic. We decided that that was a very good direction for the company, because one of the strengths of chips is that they’re very scalable. So, if you’re doing a diagnostic that’s used in a central lab, the scalability is not very important, because actually, there are not that many central labs. So, the most important thing is that the cost for like, a single test is very low, but the machine can actually be very big and expensive. But the ability to make it a consumer electronic is very valuable if you’re going into the home.
And we were already kind of thinking, “Okay, this is would be super exciting. You know, putting silicon photonics into every home, potentially even onto every arm,” right? Like so going to all the way to something like a continuous glucose monitor, but for proteins. And then during COVID it just became clear that there’s going to be a device in everyone’s home at some point. There’s so much momentum behind that, and so we committed fully to, like the at-home use case, and just making something super low cost. And that really plays to the strengths of silicon photonics.
A lot of the previous ideas for like, improvements in blood testing really centered around a faster result at the doctor’s office, which doesn’t really change the paradigm that much. We were kind of excited about, okay. Everybody we know kind of hates their experience with healthcare. What is one way that would change? The major way to change that would be to have testing at home and enable all these other like workflows that people want to do, like telemedicine and tracking yourself, and there, you know, a myriad of things that you can do if you have a home blood test. And so that’s how we ended up working on what we’re working on for the last four years at SiPhox.
Camille Morhardt 08:24
So Srini, why is Intel Capital interested in this?
Srini Ananth 08:27
What I had seen, I’ve spent about six, seven years we made investments in photonics is seen the applications in photonics driving advances in telecom, driving advances in data communications. And so, when I met SiPhox, it was a very cool application. I hadn’t thought about the healthcare application. I hadn’t thought about the diagnostics application. So, when I met them, out of YC, that was very intriguing that they brought a concept, which is, you know, miniaturization and sort of economies of scale that come from going into semiconductor processes and then bringing that to a very different field that doesn’t rely on miniaturization, right? There have been companies that have built laptop devices, or lab-top devices, right–there are large devices that sit in point of care facilities that can do pretty gold standard tests, but nothing of this type exists at home.
At home, the best you get in terms of biomarkers are your CGM, continuous glucose monitors, and you have digital biomarkers with Apple Watch and all of your Garmin and other devices. So, I think what we saw in that was the bringing that the biomarkers, and bringing that digital quantification to the home, and bringing the same gold standard to the home at a price point that consumers can afford. And what that does is, if you think about the way you manage your health care, you go once a year, you get a physical, sometimes you get the lab referral. Almost a third of people don’t even go to the labs. Don’t even get their test. Another third of the population, probably gets the test, but doesn’t care about the results that they get, because there’s no good way to track these results. They sit in a database somewhere in a Labcorp or a Quest, and then you don’t refer to it until a year.
So, what we saw with SiPhox was application of silicon photonics to bring a consumer-grade device at a consumer grade price point, but with all the validity and the accuracy of a lab test and giving you information about your hormonal biomarkers, about your health biomarkers, all those things that you would care about and you would constantly want to monitor.
And with Mike and Diedrik, we saw a team that had the diversity of experience, that brought the experience with photonics. And so, for us, this was a very logical progression to “here’s a new application.” We understand the photonic space, and there are ways that Intel Capital and Intel can help. And so that is the other angle where Intel Capital always looks at, how can we go beyond capital in terms of materially helping the company? Can we advise them? Can we provide them the components? We’ve had a great partnership with Mike and Diedrik here, and we really look forward to the application and what SiPhox can bring to the home.
I think there are applications in sort of chronic patient healthcare monitoring, if you think about companies like BioPharmac and others who have been trying to provide like patient monitoring at home, the SiPhox device can be something that provides that data points that a medical professional can use to monitor patients with chronic healthcare conditions. So that’s sort of our thesis. We are very selective in our healthcare investments, and so when we find teams and opportunities like this, we are very excited.
Camille Morhardt 11:18
Now, people are familiar with at-home tests, including COVID or at-home pregnancy tests, just over-the-counter purchases. I understand this is different because it’s using silicon photonics, but can you explain what it is you’re doing that’s more nuanced than providing these simple singular kinds of tests?
Michael Dubrovsky 11:36
So, there’s a difference between, like, quantitative tests and on/off tests. So, like, your pregnancy test just tells you you’re pregnant. It doesn’t really tell you, like, how what does how pregnant you are, right? And the same with COVID, if you’re measuring something, that really the number matters–and so, for example, you know, if the numbers 20% higher, you have a problem, if it’s 20% lower, you’re fine–that’s where it becomes an order of magnitude or two orders of magnitude more complicated to provide that test accurately. And that’s why today, in the home, you really only have glucose. There are a few others, but really what’s available broadly is glucose monitoring. And glucose is the highest concentration thing in your blood, essentially. So what people were able to do is to use electrochemical sensors, which are like electrodes with some enzymes on them, to get quantitative glucose to work in the home in a very low-cost form factor. And that’s the glucometer. That’s the continuous glucose monitor that people were in their arms. Unfortunately, that approach doesn’t really scale down.
So, once you get to like proteins and hormones, which are lower in concentration, it’s not possible to use that approach. Every single test that you get, like blood tests that you get for proteins and hormones, is run on a large optical instrument. And so really, that’s why the leverage is in miniaturizing the optics. A lot of what’s been done over the years is trying to get that electrochemical readout to work for proteins and hormones, and it hasn’t been successful, just because the physics don’t work well; and that’s why there’s no electric, you know, electronic readout. Really, in the lab, everybody’s using optics in the large lab tools.
Camille Morhardt 13:08
So, we already talked about the fact that using silicon photonics in testing is a somewhat novel approach; but as I understand it, your way of market testing your at-home test device was also somewhat novel. Would you talk more about that?
Michael Dubrovsky 13:22
Yeah, we came into this as outsiders, so we weren’t coming from Roche or Abbott or, you know, there’s just a few players in the industry for blood testing and this kind of thing. And we really try to think from first principles, like, how do you enter this market successfully? There’s like a graveyard of various devices that people have built, but mostly for point of care, not for the home. But there really a lot have been various attempts, and some of them are successful. There are also some really strong successes, but it’s definitely not a high hit rate space. And we tried to understand why, you know, and what we could do differently. And so, trying to understand what is that going to look like? Who will be the early adopters? Who’s going to come next? And so on, I think, is very important. And we weren’t really satisfied with the standard process of doing market discovery for diagnostics.
And so, what we did is we this was a very difficult decision, because it’s an enormous undertaking, but we essentially launched a preview version of our product. We took off-the-shelf technology and built what’s called a mail-in test–it’s a card that preserves the sample, and you mail it back to a central lab and they run tests on it. So, these tests already exist in the market. We basically just built a souped-up version of it, so you would get, you can go on, like, one of the standard companies is like, Let’s Get Checked, there’s a few others. So, they’ll sell, typically, a couple of biomarkers in a single test. So, we decided, “okay, let’s do a preview of what would it look like in a world where you can get, you know, 17 of these in a single test”, which is what we’re trying to do with the home device.
And we launched an entire business around it. Initially, it was really market research, like trying to understand where are the use cases? What do people want to buy? Because oftentimes, if someone’s willing to pay for something, it’s just worth ten or 100 times more in terms of information, than if they say, “yeah, like, in five years, I would use this,” it’s not very strong. So, it turned out to be pretty successful. So, we grew it into kind of a standalone, relatively large business with 1000s and 1000s of users, and we’re still learning a lot from it. One of the things that it allowed us to do is really hone in on which biomarkers we want to target first through our approvals, because we can see like where the demand is. We also have real data from tons and tons of testing, where we see which biomarkers are actually valuable in a clinical sense. And so, ok if people are using these biomarkers to optimize their health or to, you know, get early warning signs of chronic disease and so on, like, which ones are actually useful? And so, it’s taking the literature and kind of recommendations from experts and actually comparing them to empirically what are we seeing when we test people.
It’s definitely not the typical way of doing it. But so far, we think it’s actually been really instrumental for us in forming our plan for how to go to market with the hardware. And it’s also been a great way for us to just engage the healthcare industry, because I think it’s often very difficult if you’re making hardware, as engineers, how do you engage with the actual healthcare establishment, or researchers and so on, because your product is several years away, always. So, it’s having something to engage around that’s right here, right now. Like, “okay, do you want to use this now? If not, why not?” And that really creates the valuable conversations that kind of allowed us to make strategic decisions about what to commercialize.
Camille Morhardt 16:29
So, I think there’s an interesting aspect in this conversation, as well, right now around when you’re collecting data that could be linked directly to me—either blood data, biomarkers or DNA. What is the risk to my privacy? Or what is the risk to cyber-hacking or biohacking that could stem down the road from that, especially as we think of things like gene editing tools and AI kind of joining forces in the future? What are you doing to address these kinds of concerns?
Michael Dubrovsky 17:00
I think DNA is a lot more risky than blood testing, because blood testing, all the markers are pretty generic, you can’t identify somebody based on a blood test, generally. So, I think with DNA, it’s a lot more fraught. Somebody has my DNA and they know so I think there are some real concerns there. I think with blood testing, it’s less serious. Of course, at some point it does make sense to price insurance based on your blood biomarkers, right? But I think that’s really should be, like an opt-in thing, like there are some insurances which price car insurance based on your driving habits, right? So, I think from that perspective, there’s a positive possibility of being rewarded for staying healthy and so on, like, in a data driven way. I think in general, the status quo with healthcare is so bad–cost versus performance, I would say, like, that curve is not going well. So, I think it’s clear that at some point there’s going to be a lot of AI plus technology improving this situation. And so, I think, really, the data story is very positive.
I think the privacy part of it, there are a lot of laws around this, and they’re very strict. I think that privacy is like what people think about a lot, but I think the really important thing is, like, if we don’t start using the data and we don’t start automating this, it’s only going to get worse. I mean, some interesting examples are, if you take certain specialties for doctors, the number of patients is increasing while the number of doctors is decreasing because they’re retiring. So, there’s just like this completely upside down situation. So you basically need to go in the direction where you take, let’s say, the top 20% of doctors and just scale them up, right? So, make the best people able to see way more patients in some way, right? Even if it’s kind of several layers removed and so on. But I think you can also see this in terms of lifespan. So, like, the average American lifespan has plateaued, is maybe even going down, but if you look at, like, top 1% earner’s lifespan, that’s still going up, right?
So, it’s not that we’re getting worse at medicine. We’re just getting worse at providing it at scale. So, I think in many ways, the data story is like the opposite. It’s not the scary story of like, “I’m gonna lose my privacy or whatever.” I think that’s all solvable, and there’s a lot of concern about it, so it’ll probably be taken care of reasonably well. The question is, are we going to be able to leverage all this data to actually improve the situation? I’m more afraid of that not happening. I think it’s absolutely worth, like, the minor privacy risks to actually make, like, a fundamental change to the way healthcare is delivered. And so, I’m pretty optimistic about it. I would say.
Camille Morhardt 19:31
Any final thoughts, Srini?
Srini Ananth 19:34
When I look at it, it’s risk versus the opportunities and the benefits that you can get from it. The risk is, of course, yes, what if I lose my biomarkers? or my biomarkers are leaked and someone else uses it? But I look at our healthcare system is broken. We all accept that. We live with that. But you have so many different systems. You have two or three large EHR companies, and our records are all siloed and locked in. So, I can’t really carry my records from one place to another. They make it so difficult. I have no trends. I’m not able to derive any insights from all of these digital data that identify me, and so when I don’t get any benefit from it, it’s great that it’s all private and it’s all locked up, but I really don’t get any benefit from it. And in this day and age, when we’re making all these advances, why aren’t we able to see those benefits?
I’ll even take the case of in the doctor’s office, we complain about the doctor doesn’t have time to spend with the patients because they’re taking records all the time. They’re typing in notes. And we’ve been looking at companies that can do transcription, for example, within the doctor’s office. But even things like that, you can bring in AI and you can make their life easier. So, if they can spend more time with the patients, and you can entrust some of those tasks to AI, we can do a lot better job. And so, this is where open standards, around interoperability, around exchanging data between your specialist and your primary care provider, between your diagnostics provider and your primary care provider, all of those has to happen.
And if you’re someone who uses the Apple Watch for example, in the Apple ecosystem, you’re starting to see some of those integrations where the apps that you use for your fitness and your healthcare apps are tying into the Apple Watch ecosystem. And of course, people trust Apple because of the privacy element, and so you entrust this company with a lot of your data. You give them the permission, and you see some value from it. Ultimately, if you don’t see value you wouldn’t care. So, I really think it comes down to measuring what is it that you can benefit from it? And I think open standards in the past has helped with privacy, sharing the right interface, sharing the right standard format, and adhering to the same security and privacy standards, that ultimately helps.
Camille Morhardt 21:38
Mike Dubrovsky, Co-Founder of SiPhox and Srini Ananth, Managing Partner at Intel Capita. Thank you very much for today’s conversation. I found it very interesting learning about photonics but also the future of personalized medicine and on-going health monitoring.
Srini Ananth 21:56
Thank you, Camille,
Michael Dubrovsky 21:57
Thank you very much.
