MetaNovas AI

A lab where the first experiment happens on a screen

Somewhere between a research paper and a bathroom shelf sits MetaNovas. On any given morning, its software is reading - chewing through millions of scientific documents, mapping the tangled relationships between a protein, a pathway, and a wrinkle. By the time a human scientist arrives, the machine has already proposed a few molecules worth caring about, argued why they might work, and quietly discarded the ones that won't. The pipettes come later. Sometimes much later.

That is the bet MetaNovas (legally MetaNovas Biotech) makes every day: that the most expensive part of discovery is not the lab work, but choosing what to put in the lab in the first place. It is a small company - around a dozen people - with an outsized claim. It wants to design medicine, peptides and beauty actives with computation first, biology second.

Discovery is mostly a record of things that didn't work

Here is the unglamorous truth the industry rarely puts on a slide: most drug and ingredient candidates fail. They fail in screening, they fail in the lab, they fail in people. Each failure arrives expensive and late, after months of synthesis and assays. The traditional answer has been to fail faster by spending more. MetaNovas proposed a different one - fail earlier, where failure is cheap and lives only inside a model.

The founders had seen the slow machinery up close. Biology generates more literature than any team can read, more data than any spreadsheet can hold, and more dead ends than anyone likes to admit. The knowledge to make a better decision usually exists. It is simply scattered across papers nobody has time to connect.

A Stanford dropout and an MIT physicist walk into biotech

MetaNovas was founded in 2021 by Meijie Wang and Lun Yu - a pairing that reads almost like a setup for a joke, except the punchline is a working platform. Wang, the CEO, is a Stanford dropout who had worked on AI infrastructure at NVIDIA and R&D software at Olympus. Yu, the CTO, holds a PhD in Nuclear Science and Engineering from MIT and spent years as a senior data scientist steering large digital-health products at Optum, Rally Health and Aetna.

Neither came from the classic biotech mold, which is rather the point. They looked at drug and ingredient discovery the way engineers look at any slow, lossy pipeline - as something that could be instrumented, modeled and front-loaded. Their wager was that an AI which understood biology well enough to justify its reasoning could move the decisive choices upstream, before the expensive part begins.

Four engines, all named Meta

The platform is less a single product than a relay team. Each engine hands off to the next, turning a wall of literature into a short list of molecules worth synthesizing. The naming is unsubtle - everything starts with “Meta” - but the division of labor is clean.

Stacked together, they support a discovery loop with four moves: structure the evidence, generate candidates under real constraints, screen them predictively for whether they can actually be made, then hand survivors to the wet lab for mechanism-anchored validation. The model proposes; biology disposes; the model learns. Applied to drugs, it hunts novel targets and repurposes existing compounds for new indications. Applied to beauty, it engineers peptides aimed at the biology of skin aging - the work behind Yu's 2026 Paris talk, “Reprogramming Youth.”

Brands on your shelf, numbers on a deck

A discovery platform is only as convincing as the people willing to use it. MetaNovas reports strategic relationships with a roster of consumer-health and beauty heavyweights - among them L'Oreal, Beiersdorf, Haleon, Takeda and Unilever - the sort of names whose products already sit in medicine cabinets worldwide. For a roughly twelve-person team, that is a notable set of doors to have opened.

Fewer good ideas dying in the lab

Strip away the platform names and the mission is almost stubbornly practical: shorten the slow, costly, failure-prone front end of biology so that more promising molecules survive long enough to reach people. MetaNovas frames itself as a biopharma company at heart that happens to also make clinical-grade products for everyday wellness - drugs and disease targets on one side, cosmeceuticals, nutraceuticals and functional foods on the other.

It is a dual identity, and the company seems comfortable with the tension. The same engines that hunt disease targets also design the peptide in a serum. The science doesn't much care whether the end product is regulated as a drug or sold as a cream. The discipline is the same: read everything, reason about mechanism, predict before you build.

The bench is moving to the screen

The wider field is converging on this idea - Insilico Medicine, Recursion and others are all variations on the theme that biology can be reasoned about, not just experimented on. MetaNovas's particular angle is breadth: a single reasoning stack pointed at both medicine and the far larger, faster-moving world of beauty and consumer health, where it can ship, learn and iterate without waiting a decade for a clinical trial.

If the thesis holds, the interesting consequence is cultural as much as scientific. Discovery stops being a story told mostly in failures and starts being one told in better first guesses. That is a quieter revolution than the marketing usually allows - but a real one.

Return to that morning lab. The software has been reading all night again. By the time the scientists arrive, the candidate list is shorter, the reasoning is attached, and the obvious mistakes are already gone - made and unmade inside a model that costs nothing to be wrong. The pipettes still come out. They just come out later, and for fewer molecules, and with better odds.

That is what MetaNovas is changing about the scene it walked into: not the existence of failure, but its timing and its price. The pipettes wait. The reading never stops.