3D illustration of a molecular model of an antibody. [Adobe Stock]
The aim is to use proprietary, novel immune data to drive AI-powered drug discovery. Parallel Bio’s website notes that the “massive scale of organoid populations generates novel immune training data at unprecedented rates.” The system can then validate AI-discovered candidates in vitro in a large population.
Gesher envisions a future where biologists focus more on designing experiments rather than executing them. In this scenario, a researcher might say, “I want to run this experiment. I need 50 different human models from different donors. I want to show the results of both control and treatment branches of the experiment,” Gesher said. These parameters could be input into a system that would then carry out the experiment autonomously. As Gesher puts it, “Your whole frame rate increases, your ability to do different kinds of experiments changes your relationship to the labs.” This approach doesn’t just speed up existing processes; it opens up entirely new avenues for scientific inquiry.
From methane hunting to organoids
Ari Gesher
Gesher’s arrival on the biotech scene isn’t conventional. His previous venture, Kairos Aerospace (now Insight M), tackled a pressing environmental challenge: methane leaks from oil fields. “15% of global warming is caused by fugitive emissions on oil fields,” Gesher said. But identifying methane leaks can be like looking for a proverbial needle in a haystack. Picture a hundred square miles dotted with a million components, where just a handful of faulty pieces contribute to the vast majority of harmful methane release. Gesher helped industrialize the methane detection process at Kairos, turning a complex scientific challenge into a scalable solution. .
The shift from combating climate change to biotech might seem like a leap, but for Gesher, it represented another opportunity to apply his expertise in scaling complex technologies. “I took some time off and I was just interested in talking to people doing interesting work. It wasn’t even a focused job search,” Gesher said. “But I got on the phone with [Parallel Bio founders] Robert [DiFazio] and Juliana [Hilliard], and they explained to me what you could do with immune organoids. Literally 20 minutes into the conversation, part of my brain that lit up. I was like, ‘Cool. We’re going to do this.'”
Developing a platform based on the human immune system
At Parallel Bio, a Cambridge, Massachusetts-based biotech startup founded in 2021, researchers are developing a platform that replicates the human immune system in a dish for drug discovery and development. The company aims to flip the pharma industry’s high failure rate “on its head,” where the industry can succeed more often than it fails. “That’s an iteration of modern medicine that we’ve never seen,” he said.
While improving pharma’s R&D calculus has proved challenging, that could change. Gesher draws parallels from the history of technological transformations. “Take the first industrial revolution, which was unlocking external sources of energy to do work for us,” he said. “It’s interesting how long it takes for these things to become radically transformative, but when they hit, they hit.” For instance, the first practical steam engine was invented in 1712, improved by James Watt in 1776, but the major impacts of industrialization weren’t felt until the 19th century.
A tech blueprint for biotech
Gesher points to computing as a source for future shifts in biotech. “We’ve been through 25 years of computing revolution, which really had its ‘steam engine moment’ at the end of World War II with the first digital computers,” he explained. “Just in the last 25 years, we’ve watched computing deployed into everything. We now have screwdrivers with microprocessors and gyroscopes – that level of ubiquity was unimaginable not long ago.”
Consider that the Apple II computer used the MOS Technology 6502 microprocessor, an 8-bit chip introduced in 1975 with 4,528 transistors. Contrast that with the Apple M2 chip found in the various MacBooks and iPads which has 20 billion transistors. The M2 Ultra has 134 billion.
At the same time, advances in AI-based drug discovery models with support from industry leaders like NVIDIA point to a similar exponential growth in computational power applied to drug development and screening processes. Added to that are advances in lab automation that can result in entirely new approaches to conducting basic biological research. “It’s a whole level of not being able to imagine what the world is gonna look like on the other side,” Gesher said.
Tackling drug discovery inefficiency
The pharma industry, for all its advances, faces a fundamental challenge: inefficiency. “I think if we look at drug discovery writ large, what’s missing [in early-stage development] is testing in humans,” Gesher said. “Everybody’s making these sort of really long-shot bets, and we’re like, ‘Let’s see how it goes once it gets to people.'”
Animal models compound the problem. “Organoids have the potential to really change the entire pipeline structure in pharma because you get to do much earlier testing in very high fidelity, what are essentially human models,” Ari explained. “They’re not actual humans, but they behave much more like humans than an animal model would.”
Organoids also offer a unique advantage over traditional models in terms of replicability. “One of the problems with testing in organisms, be it people or lab animals, is that you can never run the same experiment twice,” Ari pointed out. “You actually can’t do a controlled experiment on the same organism. Here, you can instantiate the same organoid multiple times and treat it differently. Now the biology is held constant, and you don’t have that same variance.”
The vision of a ‘lights-out’ lab
Ari’s vision for Parallel Bio extends far beyond simply using organoids—he’s building a “lights-out” lab where automation takes center stage. “I’m interested in building this kind of ‘lights-out’ lab from the beginning,” Ari explained. “I want all the data to be able to come off of it. I’m thinking about that design from the beginning.”
But this isn’t just about replacing human hands with robotic ones. “I’m not starting with ‘How do I get my biologist to be more efficient?'” Ari said. “I’m using our biologists to prototype what the scientific processes are, then we’re going to build automation.”
Reimagining scientific discovery
“The place that I’m coming from,” Ari explained, “with a somewhat unique perspective from computing, is that I’m interested in … not starting with efficiency, but with desired outcomes. Then we build the automation to achieve those outcomes.”
This approach has significant implications. It means that instead of simply automating existing tasks to make them faster, the lab itself can be designed around achieving specific biological goals, often in ways impossible with manual methods.
“When you’ve built this really automated lab, you actually have the ability to do things that you couldn’t really do before,” Ari said. It’s not just about speed or cost; it’s about pushing the boundaries of what’s scientifically possible. “It lets you design a different kind of experiment to get a lot different signal or a higher quality signal than you could otherwise.”
This “lights-out” lab, then, becomes more than just a collection of machines—it transforms into a powerful engine for scientific discovery, poised to usher in a new era of biological research. “I see biology doing the same thing as computing did,” Gesher said. “Only it’s like it’s the late 50s in biology by comparison to computing. This revolution hasn’t even picked up steam yet…. Once you bring computing into biology in a very real way, you’re going to see all these different architectures and design patterns come in.”
Good Job dad
This is my brother
That’s my dad!