Interview with Joseph Geraci
Co-Founder & CEO @ NetraMark

NetraMark is a tech company specializing in creating machine learning methods that can deal with the challenges that many modern methods cannot. Deep Learning and tree-based methods are very powerful but there are many instances where they are not ideal. Specifically, we have created a novel mathematical paradigm for reinforcement learning that is capable of dealing with small data, capable of extracting insights from data that has a lot of ‘unknown unknowns’ hidden within, and it can explain itself. We provide an augmented intelligence platform that amplifies innovation. One of our focuses is in Personalize Medicine. Drugs don't work for everyone and may not be safe for certain individuals and  NetraMark Corp is bringing precision medicine to the market. Our efforts can bring effective and safe drugs to millions of people who would otherwise not get them. Our NetraHealthAtlas is an AI that uses biochemistry and patient data to understand how disease states should be defined, and it offers potential ways to treat them. These new disease states are being defined by our system and the underlying molecular circuitry learned. What makes all this possible is a beautiful mathematical paradigm developed in-house that captures the many possible vantages of a data set, like pages bound into a book. We call this process NetraAI, from the Sanskrit word netra meaning ‘to see’.
Joseph, what were your main motivations behind founding NetraMark? What was it like deciding to start a medtech company of your own?
So, I have a unique background. I have a doctorate in mathematical physics, but I quickly jumped into oncology out of interest. So, I did a postdoc in oncology, I did a postdoc in machine learning and I did another postdoc in neuropsychiatry. And so, from this vantage, I started realizing that there's a lot of structure to disease - disease states that the classical paradigm have been unable to reveal, but that this new paradigm fueled by mathematics would be able to. And so, I was approached by several people who were interested in my approach and to start businesses. And finally, I found the right partner. So, what I did was, I ended up leaving my academic positions and raising money. And, you know, the main motivation for doing this was specifically to do things my way from this perspective of personalized medicine. Is there a way to define disease in a way that can really speed up clinical trials and to begin to reconnoiter disease states in an efficient way. And so, that's how we started and what was it like. It was stressful and terrifying, but very rewarding. It's the best thing I've ever done.
How do you balance your academic career and your work with NetraMark? How do the two spheres of your career intersect?
Well, there's actually very little balance between academia and NetraMark, you know, the demands of a startup are brutal as you can imagine. So, I mean between, you know, balancing and coordinating the tech team, developing new technologies, testing, dealing with clients, you know, it's very time-consuming. However, I'm an assistant professor at Queen's University in Ontario. And so, we are actively engaged in doing academic research. And I, you know, I try to do as much as possible. But, the real opportunities come when we're able to do these studies in depression, and ADHD, and bipolar disorder, certain cancers. And we actually use the new technologies to rigorously evaluate how they do with these data from these academic studies.
What are some of the main misconceptions the public has regarding the potential for data in the medical industry? How do you help to dispel some of these misconceptions with NetraMark?
I guess in general most people are afraid to share their data, but they don't realize how important each person's data is and the reason is this: originally the way we have created our current classification systems like when a doctor says breast cancer or Alzheimer's - these are very clumpy definitions that have been extracted through observation and a lot of hard work. But, now with the advent of modern computer science, and machine learning and mathematics, we're able to actually utilize your data to increase the resolution with which we actually understand the disease states. And so, what this means is that we'll be able to personalize a lot better. Now, at NetraMark what we're doing is, we're, and I'll expand on this in another question, but we're actually creating a new set of definitions for disease based on the way people come together in these datasets. So, I think that's what we're doing to dispel this and, you know, and we love talking about it…getting people to understand the actual potential behind these data. And, data that goes beyond the popular types of genetic data that you know certain companies are collecting from people. There's other things that we can capture as well. And so, we're excited to be engaging in that respect.
Since you founded the company in 2015, how has the company changed or shifted its service offering? What have been the main growth benchmarks you've reached with the company thus far?
So, in 2015 what we actually had was just a next-generation machine learning algorithm that was different from the usual tree-based methods and from deep neural networks. And, what we had was a technology based on a different set of mathematics that was really great at seeing disease states at a much higher resolution. And, this allowed us to do two things well…with increased capacity, I should say. We're able to understand placebo response and predict placebo non-response. And, we were able to identify subpopulations that actually responded to drug. This allowed us to go to pharmaceutical companies and say listen you're having a difficulty with your Phase 3 trial or you're about to commence something that you expect to be difficult. We can help you isolate those groups of patients for which your drug will work. And, we can help you with placebo response. However, as we progressed what happened was a pharmaceutical company came to us who was focused on aging and they became interested in using our technology to help them resurrect failed drugs. So, take drugs that have failed clinical trials, but use our technology to resurrect them to be able to do another trial with them. Now, we formed this joint venture which was a major milestone for us. Now, in addition to that, another thing that happened, you know, that could be considered an advancement in the way we conduct our business, is we have actually found that our technologies are useful in other verticals. So, we have now commenced a pilot in the smart building space and we also are actively involved in the quantum computation vertical as well.
What's next for your work with NetraMark? What are the main technical innovations, partnerships and projects you'll be focusing on in 2019?
So, NetraMark is a technology company which focuses on machine learning. And, we have solutions in the smart building space. And, that has been captured in a joint venture called Netra IOT and, we have a pilot project in a building in downtown Toronto. And, we have solutions in the pharmaceutical space and, we have a joint venture called NetraPharma, and that's focused in the pharmaceutical space. And, we also have novel solutions that we are developing in the quantum machine learning space as well. Now, our approach is to embrace classical techniques like deep neural networks and boosted trees and so forth, and also to create next-generation techniques, like DeepCrush, which we're very excited about. We actually have put forth an online cloud-based version of the DeepCrush algorithm for people, for pharmaceutical companies specifically, to engage and play with their patient populations. But, in addition to all this, what I'm most excited about in 2019 is the development of this A.I. that works while we sleep. It's called The NetraHealthAtlas and what it does is essentially redefine disease. We put disease states in there so, these are databases not necessarily large and that's one of our advantages. We don't need very big data sets because of our next generation technology and what it does is, it digests this patient population data and suggests novel ways of looking at the substructure of the disease. And, it infers the pathways that are involved, and it tells us possible drugs that could be repurposed for those diseases. And, we're very excited to be applying this to the aging space and to cancer and to neurodegeneration. Thank you.