The idea behind Onepot AI emerged from a shared frustration experienced by co-founders Daniil Boiko and Andrei Tyrin: in modern drug discovery, the most innovative therapeutic concepts often stall long before reaching the lab bench.
“Breakthrough ideas weren’t dying because of biology — they were dying because no one could make the molecules,” Boiko explained.
Small-molecule synthesis — essentially assembling new compounds through controlled chemical reactions — is a bottleneck that can halt projects before scientists ever test their potential. Boiko, who researches AI applications in chemistry at Carnegie Mellon, saw promising drug candidates routinely abandoned simply because synthesizing them appeared too complex.
Tyrin, who previously worked on computational drug-discovery systems, observed a similar mismatch. “Models could generate thousands of viable compounds in hours, but the chemistry teams took months to catch up,” he said. Both founders realized that while AI was accelerating ideation, the ability to actually make compounds hadn’t kept pace.
Their conclusion: the world was investing heavily in molecular design while neglecting the harder and more foundational challenge — rebuilding chemical synthesis.
Advertisement
A New Approach to Chemical Synthesis
In 2024, Boiko and Tyrin launched Onepot AI, which operates a dedicated small-molecule synthesis facility (POT-1) and an AI chemist named Phil that analyzes experimental data and proposes synthesis strategies. Early biotech and pharma partners are already using the system.
Onepot AI emerged from stealth this week with $13 million in funding, spanning pre-seed capital and a seed round led by Fifty Years. Khosla Ventures, Speedinvest, Wojciech Zaremba, and Jeff Dean also participated.
Today, most drug developers either hire internal synthetic chemistry teams or outsource projects to overseas CROs — a slow, costly process. A single compound can require months of bench work and thousands of dollars in labor.
“The real constraint isn’t testing compounds; it’s producing them,” Tyrin said. “We want to shrink that cycle from months to days.”
How Onepot Works
Onepot AI offers a catalog of compounds it can synthesize. Customers select what they need, and Onepot AI manufactures the molecules — delivering them as dry material or in ready-to-use solutions for downstream experiments.
Behind the scenes, the platform is powered by:
- a continuously expanding internal laboratory,
- automated data capture down to every reagent, temperature shift, and procedural step,
- LLM-based agents trained not just on literature, but on raw experimental data produced within the lab.
“We don’t lose any experimental detail,” Tyrin said. “That makes every reaction reproducible, even a decade later.”
This real-world data gives Onepot’s AI models an advantage: they learn from actual chemistry, not just published reactions.
Boiko and Tyrin see their main competitors as traditional synthesis powerhouses like WuXi AppTec and Enamine, but argue that they can move significantly faster by closing the loop between experimentation and AI-driven synthesis planning.
Funding, Expansion, and Next Steps
Boiko described fundraising as intense but energizing — their lead investor meeting unexpectedly turned into a multi-hour deep dive on how to industrialize synthesis from scratch.
With the new capital, Onepot plans to:
- open a second lab in San Francisco,
- onboard more biopharma customers,
- expand its synthesis workflows, and
- continue growing its compound-discovery engine.
The long-term ambition is bold: to redefine the limits of drug design by venturing into chemical space previously deemed too difficult or too unusual to attempt.
“You’re not merely accelerating drug discovery,” Boiko said. “You’re expanding the boundaries of what kinds of medicines and materials are even possible.”
