The scientist who traded machines for microbes

Somewhere in the middle of a PhD, most people put their head down and finish the thing they started. Will Cao did the opposite. She had been training to engineer machines. She decided, instead, to engineer bacteria - to open up the cell, rewrite what it does, and turn living systems into tools. That single swap, from the mechanical to the biological, is the thread that runs through everything she has built since.

Today Cao is the co-founder and CEO of Pando Bioscience, a company that came out of Y Combinator's Winter 2023 batch and now works out of Watertown, Massachusetts. The pitch is small enough to fit on a t-shirt and stubborn enough to be a whole business plan: life is too short for slow enzymes.

The boring problem nobody was shouting about

Enzymes are the invisible workhorses of modern medicine. They stitch molecules together, break them apart, and make it possible to manufacture drugs at scale. When an enzyme is good - fast, stable, precise - a manufacturing line hums. When it is slow or fragile, costs pile up quietly, batch after batch, and nobody makes a headline out of it.

That is exactly the kind of problem Cao is drawn to. Not the loud crisis, but the accepted inefficiency that everyone has learned to live with. Designing a better enzyme has traditionally meant months of trial and error, balancing one property against another, and hoping the winner survives real conditions. Pando's whole reason for existing is to make that process fast, cheap, and repeatable.

Those numbers describe Pando's ultra-high-throughput screening platform, which the company says evaluates roughly a thousand times more enzyme variants than conventional methods - and does it faster and cheaper. The screening is the engine. The generative AI sitting on top of it is the brain.

A loop that gets smarter every round

The elegant part of Pando's approach is not that AI replaces the lab. It is that the AI and the lab feed each other. A model proposes enzyme designs - specific mutations meant to improve several properties at once. The screening platform builds and tests them under real conditions, turning messy biological signals into clean, quantifiable data. That data goes back into the model, which proposes better designs next time. Round and round, each pass sharper than the last.

The payoff shows up in the hit rate. Pando says validated enzyme hits jump roughly tenfold between the first and second optimization round, and that AI-guided improvements consistently beat random mutagenesis. In a field where a single design cycle could once eat a season, compressing the loop to about a month is not a convenience. It is the difference between a therapy being affordable and being out of reach.

The road to Watertown

Cao's path reads like someone collecting exactly the right tools before building something of her own. She did research at Duke University and was tapped as an Advance Into Consulting invitee at Bain & Company. Then, in 2018, she joined Ginkgo Bioworks - the synthetic biology unicorn - as an early employee, working as a biological engineer and later a senior design engineer.

Along the way she published first-author scientific papers in Cell and Nature Biotechnology, the kind of results most scientists would happily build a career around. Cao used them as a foundation instead. In 2022 she co-founded Pando with Yang Wang, who spent six years at Harvard Medical School before the two of them crossed paths at Ginkgo and started thinking about enzymes as a business.

The polymerase with a nickname

Pando does not only sharpen other companies' enzymes. It is building its own. The flagship is a high-fidelity polymerase the team nicknamed Saiyan-Phi29, engineered to surpass market leaders on thermostability and activity while keeping low GC bias across several genomes. A polymerase is the enzyme that copies DNA, and a better one ripples out into diagnostics, sequencing, and research tools. Naming it after a Dragon Ball power-up is the kind of small, human touch that tells you a real team is behind the science, not a slide deck.

Off the bench

Cao goes by Will, though her given name is Yangxiaolu. She is also a certified personal trainer, and when she is not thinking about enzymes she is kayaking or scuba diving. It fits the pattern of someone who does not accept that a person has to pick a single lane - engineer or biologist, scientist or founder, indoor or outdoor. The most interesting people rarely choose just one, and Cao seems to have decided early that she would not either.

That range shows up in how she represents the work publicly. She has taken the stage at SynBioBeta and the ChinaBio Partnering Forum, speaking about integrated platforms that design, screen, and optimize therapeutic enzymes - the practical, unglamorous machinery of turning AI promises into things that actually work in a lab.

For now, Pando is a small team - around five people - competing against far larger enzyme operations. The bet is that in modern biology, leverage comes from software and smart screening rather than sheer headcount. If Cao is right, a handful of people in Watertown feeding a smarter and smarter loop can out-engineer companies many times their size. And every slow enzyme they replace is a small argument that the mission on the t-shirt was serious all along.