The Harry Glorikian Show

Harry's guest this week is Raffi Krikorian, chief technology officer and managing director at Emerson Collective, the social change organization founded by Laurene Powell Jobs. Krikorian is the former vice president of engineering at Twitter (now X), where he was responsible for getting rid of the Fail Whale and making the company’s backend infrastructure more reliable; the former director of Uber's Advanced Technology Center in Pittsburgh, where he oversaw the launch of the world's first fleet of self-driving cars; and then the chief technology officer at the Democratic National Committee, where he helped rebuild the party's technology infrastructure after the Russian hacking debacle of 2016. At Emerson Collective, Krikorian built the technology organization, leads the development of data products, and works to upgrade the back offices of the non-profits Emerson works with. On top of all that, he recently launched a podcast called Technically Optimistic, where he’s taking a deep dive into the way AI is challenging us all to think differently about the future of work, education, policy, regulation, creativity, copyright, and many other areas. The show is a must-listen for anyone who cares about how we can build on AI to transform society for the better while minimizing the collateral damage. Harry talked with Krikorian about why he moved to Emerson Collective, why and how he started the podcast, and what he really thinks about what government should be doing to prepare for the waves of social change AI will bring.

Generative AI isn’t magic. You can’t just  sprinkle it like pixie dust over an existing project or dataset and expect wonderful things to happen automatically. In fact, just to use the data you already have, you have to you may have to invest a lot in the new infrastructure and tools needed to train a generative model. And that’s the part of the puzzle Harry focuses on in today's interview with David Buniatyan. He’s the founder of a company called ActiveLoop, which is trying to address the need for infrastructure capable of handling large-scale data for AI applications. He has a background in neuroscience from Princeton University, where he was part of a team working on reconstructing neural connectivity in mouse brains using petabyte-scale imaging data. At ActiveLoop, David has led the development of Deep Lake, a database optimized for AI and deep learning models trained on equally large datasets. He says the company’s goal is to take over the boring stuff. That means removing the burden of data management from scientists and engineers, so they can focus on the bigger questions—like making sure their models are training on the right data—and ultimately innovate faster.

Harry's guests this week are Frank Cheng, CEO of UK-based Caristo Diagnostics, and Keith Channon, Caristo's co-founder and chief medical officer. Under their leadership, Caristo has introduced an AI-based test called CariHeart that applies machine learning to the data in a three-dimensional CT scan of the heart. It looks for signs of inflammation in the fat tissue around the major coronary arteries—a risk factor that's often overlooked because it isn't always accompanied by plaque or narrowing of the arteries. Doctors can use that information to decide whether a patient needs to take a cholesterol-lowering drug like a statin or an anti-inflammatory drug like colchicine. Caristo’s test is being used on an experimental basis in the UK, and it hasn’t yet been approved for use in the US. But it’s a leading example of the way AI, put together with fundamental advances in our understanding of human biology, is really beginning to change the practice of medicine.

If you believe that computation will help companies get better at developing new drugs, then what specific kind of computation and software should you invest in? Quantum chemistry simulations? Molecular dynamics simulations? Generative AI models? Harry's guests this week, Garegin Papaoian and Michael Antonov, lead a company called Deep Origin that's taking an all-of-the-above approach. The company’s philosophy is that physics-based modeling by itself won’t be enough to build a powerful drug discovery engine. But neither will generative AI, which requires more training data than lab scientists will ever be able to provide. They think the only reasonable approach today is to combine the two, and use both physics and AI to try to get better at predicting which molecules could become effective drugs. It’s important stuff, because if Deep Origin is right, then a lot of other more specialized biotech and techbio startups could be going down the wrong path.

If you look back at all the health-tech and drug development companies Harry has hosted on the show, an interesting pattern starts to emerge: a very large number of those companies have gone on to enormous growth and success in their markets. It could be that being on the podcast is like a catapult to success—or it could be that we're pretty good at finding companies that are already on a promising trajectory. Either way, there's no better example than Concert AI. The company’s CEO, Jeff Elton, first spoke with Harry back in July of 2021. At that time, the company was already one of the leaders in gathering and analyzing broad collections of data about cancer patients involved in clinical trials for new treatments. Its specialty was, and is, going beyond the very specific endpoints measured in clinical trials and looking to electronic medical records, genome sequencing data, insurance claims data, and other sources in order to build a more comprehensive picture of cancer patients and their journeys through the healthcare system. That kind of data can be very useful to companies trying to track the performance of their drugs after they’ve reached the market, and to researchers planning new clinical trials. And since that first conversation, the company has grown by leaps and bounds. It’s taken over management of more data sources, including the massive CancerLinq database formerly maintained by the American Society of Clinical Oncology. It’s struck up partnerships with some of the leading technology startups, research centers, and drug companies working to beat cancer. And it’s leaning hard into the new wave of deep-learning AI tools and their potential to help find patterns in vast amounts of data about patients. It’s probably safe to say that ConcertAI has gathered up more data about cancer patients than any other company on the planet. And investors have been rushing to pour money into the company, on the conviction that data is going to be the key to getting more and better cancer drugs to market. That’s certainly Jeff Elton's conviction too, as you’ll hear in today's interview.

One of the most amazing successes in the battle against cancer over the last two decades has been the introduction of antibody drugs that harness the body’s own immune system to kill tumor cells. Finding those drugs may sound like a biology problem rather than a machine learning or a big-data problem. But actually, these days, it’s both. Harry's guest this week is Leonard Wossnig, who’s the chief technology officer for a UK company called LabGenius. The company uses a combination of synthetic biology, high-throughput assays, and machine learning to hunt for new drugs within a subclass of antibody medicines called T cell engagers that, loosely speaking, can grab tumor cells with one end and then grab tumor-killing T cells from the bloodstream with the other end. And Wossnig says the key to the whole thing is having the best data possible—meaning, data about their candidate T cell engagers and how specifically they bind to their targets in the lab assays. LabGenius has built an automated platform called EVA that runs experiment after experiment and uses active learning to zero in on T cell engagers with just the right ability to bind to their intended targets. One of the big takeaways from the interview is that companies that want to use AI to speed up drug discovery need the biggest, cleanest, and most consistent data sets possible.

The combination of better data and more powerful computing is helping researchers reinvent the process of discovering new drugs. Within 5-10 years, we’ll likely see a huge wave of new medicines that were either discovered or designed using AI—drugs that will finally help us get control of our most stubborn health problems, from cancer to cardiovascular disease to obesity and metabolic disorders to neurodegenerative diseases. And the biotech startups that will do most to contribute are the ones that have both proprietary data, and original ways to use AI to sift through that data. Harry's guests this week are from a startup called Pangea Bio that’s working hard on both. As Pangea's co-founder and COO, John Boghossian, and its president of AI, Sona Chandra, explain, the company specializes in gathering data from the natural world, especially data about compounds manufactured inside the cells of plants and fungi. They narrow down the possibilities by working with indigenous cultures to find the plants or mushrooms that people have already been using for centuries in traditional medicine. They've also built three separate computational platforms that filter through all that data, to single out the small molecules that have the biggest effects in the human body, especially the central nervous system.

The microbiome has been getting more and more attention from researchers and doctors now that we’re starting to have the tools we need to identify and measure all those microbes and see what they’re up to. Harry's guest this week is serial healthcare and AI entrepreneur Leo Grady, whose company Jona is on a mission is to help patients and physicians keep up with the skyrocketing amount of scientific literature about the microbiome and try to translate it into real steps people can take to improve their health. If you’re a Jona customer, you start by sending in a fecal sample. Then the company uses a large-scale gene sequencing technique called shotgun metagenomics to get a profile of all the microbes in your GI tract. Since everyone’s microbiome contains a different mix of microbes, the next step is to use large language models to sift through the published science about the microbiome and find the studies that relate to the specific bugs in your microbiome. Then the company gives patients and their doctors a report that parses out whether their microbiome makeup might be contributing to their health problems, and whether there might be any health or nutritional interventions that would help. It’s all in the early stages. And right now Jona’s test is mostly available through concierge medical services, executive health clinics, and other offices that do a lot of cash-pay tests. But Grady thinks that over the long term the service has the potential to turn the microbiome from a former black box into something closer to what he calls an “organ of data”—meaning, a part of the body that doctors can, in a sense, visualize and analyze in the same way we can use MRI and other forms of imaging to scan our other organs.

Quality control is one of those things that only a select few people pay attention to—until something goes wrong, then everyone cares. That's especially true in the drug manufacturing industry, where episodes like cross-contamination in a drug factory can shut down a production line and create instant shortages of important medicines. And if a contaminated medicines ever does get shipped out to clinics or stores, people’s lives can be at stake. So drug makers are usually pretty receptive toward any new technology that can help them detect manufacturing problems before they get out of hand. That’s the market opening that Harry's guest this week, Taylor Chartier, says she saw back in 2020, during the coronavirus pandemic. Chartier watched the stories about the Baltimore company Emergent BioSolutions, which was manufacturing vaccines for Johnson & Johnson and AstraZeneca and had to throw out millions of doses of both vaccines due to suspected cross-contamination, and thought: there has to be a better way. So she started her own company. And today her startup Modicus Prime is partnering with top pharma companies to use new machine vision and AI capabilities to catch drug manufacturing problems faster.

There’s a lot of talk out there about how artificial intelligence will change the way doctors and nurses take care of patients; you hear some of it right here on this show. But all of that still feels like a forecast rather than a present reality. When you look really closely, it’s hard to find concrete examples where AI is already helping healthcare providers make better decisions that improve patient outcomes and take costs out of the system. That’s why Harry wanted to have Nassib Chamoun on the show. Chamoun is the founder and CEO of Health Data Analytics Institute (HDAI), which has been working with a major healthcare system, Houston Methodist, to test out a working platform called HealthVision. It's a collection of AI-driven models that use huge amounts of data, both from Medicare and from Houston’s own electronic health record system, to make predictions that help doctors and administrators spend less time poring over records and data, and more time interacting with actual patients and making good clinical and management decisions.