For Daniil Boiko and Andrei Tyrin, the thought for Onepot AI got here from the similar frustration.
“The finest concepts in drug discovery have been usually blocked not by biology, however by synthesis,” Boiko instructed TechCrunch. Synthesis is the creation of latest molecules through the use of chemical reactions. It’s like a recipe or Lego items, the place small items, elements, molecules, come collectively to kind a wider puzzle image, a meals dish, a bigger molecule.
As one would possibly anticipate, it’s fairly onerous to create these small molecules that go on to construct larger ones.
For Boiko, a Ph.D. candidate learning machine studying in chemistry at Carnegie Mellon (he obtained his bachelor’s and grasp’s in natural chemistry from a college in Russia), that meant realizing that drug hunters — the scientists who oversee drug discovery and improvement — have been skipping promising concepts simply because the chemical molecules to create the medication appeared too onerous to make.
“The compounds by no means even acquired an opportunity to be examined,” Boiko instructed TechCrunch.
For Tyrin (who obtained his bachelor’s in laptop science at MIT), his time working on drug discovery computational pipelines made him understand how behind the world of drug discovery was. “The fashions may generate concepts in hours, however it may take months for the lab to catch up,” he instructed TechCrunch.
“We each noticed that the world was throwing cash into molecular design and nearly ignoring the more durable drawback of really making the molecules,” Boiko mentioned. However there was a geopolitical angle too, he continued, international provide chains have gotten susceptible, and the U.S. is getting into a commerce warfare and modern competitors, as soon as once more, with China.
Techcrunch occasion
San Francisco
|
October 13-15, 2026
“It was clear,” Boiko mentioned. “Small molecule synthesis wanted to be rebuilt from the floor up in the United States.”
Boiko and Tyrin got here collectively to create Onepot, an organization that’s dwelling to the small-molecule synthesis lab POT-1. Additionally they constructed Phil, an AI natural chemist, to assist run experimental evaluation to extend the course of of compound synthesis for their early industrial companions. These companions are biotech and pharma firms which can be at present attempting out their expertise.
On Wednesday, the firm got here out of stealth with $13 million of funding, together with pre-seed cash, and a seed spherical led by Fifty Years.
“At present, pharma and biotech firms both construct whole groups of chemists in-house or work with contract analysis organizations abroad,” Tryin mentioned of the course of for molecular synthesis. Human chemists can spend months of analysis to create even a single compound, at a value of 1000’s of {dollars}.
It entails loads of trial and error — learning numerous compounds, amassing information on organic exercise, how the drug strikes by the physique, toxicology stories, and developing with what to experiment with subsequent. “The most important limiting issue right here will not be testing these compounds, however making them in the first place,” Tyrin continued. “We goal to compress this right down to days.”
Tyrin mentioned the product is fairly easy. Onepot has a listing of molecules it will possibly make. Shoppers select which compounds it need and then Onepot’s expertise will synthesize the molecules and then ship them to the buyer so the buyer can use them in their very own experiments. (They ship bodily merchandise both as dry compounds of options in plates or vials).
The backend of the product is the place Boiko and Tyrin have enjoyable, dissecting the issues of chemical synthesis to search out out which mixtures of molecules work collectively. They constructed a lab the place they’re letting LLM brokers entry these so-called molecule recipes for coaching so the brokers may also discover out what works and doesn’t in compound constructing.
“When executing experiments in the lab, we seize each single element that goes into the course of,” Tyrin mentioned — which means tracing temperature and, basically, the elements that have been added to a combination to create compounds. “No info is misplaced, which makes experiences reproducible even when somebody decides to run them in ten years from now.”
This additionally means their brokers generate hypotheses from real-world experiments reasonably than literature information, usually mined from the web.
Boiko known as the fundraising course of “hectic,” and mentioned they met their lead investor by an intro. “What was alleged to be a brief assembly was a multi-hour whiteboard session about industrializing synthesis,” Boiko mentioned. Others in the spherical embrace Khosla Ventures, Speedinvest, OpenAI co-founder Wojciech Zaremba and Google’s Chief Scientist Jeff Dean.
The contemporary capital might be used to construct a second lab in San Francisco so the crew can tackle extra clients. It is going to additionally increase the crew and its compound discovery engine. On the service aspect, Boiko and Tyrin have a look at WuXi AppTec and Enamine as rivals.
General, Boiko and Tyrin hope to make drug discovery at the very least two occasions quicker and hopefully change the notion of what’s potential after one faucets into the “bizarre” chemistry that scientists as soon as thought was off limits.
“You’re not simply dashing up drug discovery, you’re increasing the design house for what medication and supplies will be,” Boiko mentioned. “That drug that we haven’t found but, is perhaps out there, ready for us to search out it.”
