twig bio is a London startup that builds microbial strains to manufacture ingredients used in consumer goods: sunscreen, cosmetics and personal care products. Its foundation model, Canopy, is designed to find and validate the right microbes faster than conventional methods allow. Its first product, Mycolux, is a fermented substitute for the petrochemical-derived components used in sun protection.
Foundation models in biology have largely focused on drug discovery. London-based twig bio is applying the same approach to biomanufacturing, using its model Canopy to design microbial strains that produce ingredients for consumer goods like sunscreen, cosmetics and personal care.
Canopy is trained on more than three years of proprietary experimental data generated in twig’s lab, and is designed to shorten the path from metabolic engineering strategy to a production-ready strain. Its first product, Mycolux, is a fermented alternative to petrochemical-derived ingredients used in sun protection.
Satnam Surae, twig's chief technology officer, spoke with The Infinite Loop about what proprietary experimental data changes in strain design
and what worries him most about getting a working microbe out of the lab.
'We built the data first, then the model'
TIL: A company comes to you and says, "We need a fermented substitute for material X." What do you do next?
Surae: What we do, and how we help customers, is replace chemicals used predominantly in personal care and cosmetics. We do that by leveraging Canopy, our multimodal knowledge-graph foundation model, which helps us design metabolic engineering strategies: the microbes we then build in our lab and test in fermentation systems to produce those molecules at scale. Canopy lets us do that quickly, effectively, and with high confidence that when we get to the expensive part of the process in the lab, those designs will actually work.
TIL: Most people, when they hear “foundation model” in biology, think of drug discovery. Why did you go toward bioengineering instead?
Surae: I've worked in pharma, but I've also worked in the chemicals industry, and there's a real gap in terms of specialized models for it. We lean on more generalized protein models and take inspiration from a lot of the pharma-focused ones, but we needed to build this model ourselves because nothing out there would really help us. We took the initiative: we saw the gap, we understood the space, and we've been building data in our lab for more than three years to get here. We built the data first, then the model, and now we can iterate quickly and generate more data to accelerate the model's capability.
'No one had ever taken these molecules to market'
TIL: Who are your customers?
Surae: We're selling to specialty chemicals customers and consumer packaged goods companies. We currently have three customers: one in the US, one in the UK and Europe, and one in Asia, so we're already spreading geographically. Our first product comes to market later this year — it’s a novel biomolecule to replace the chemicals in sunscreen. That's a massive opportunity, because most sunscreen chemicals are endocrine disruptors. They're environmentally unfriendly, and critically, they come from petrochemical sources, which means they're exposed to supply-chain and price volatility on top of the sustainability problem.
TIL: How did the idea of this product come about, and what did it take to get to market?
Surae: Mycolux is a formulated ingredient that goes into end products. We sell it to customers, who package it into the formulations consumers actually buy. The idea came from a gap we spotted, a real need for innovation in sun protection, with few good options on the market. We understood the science and knew it was feasible, but no one had ever taken these molecules to market, so that was the risk. The traction's been fantastic: going from nothing to a production strain in manufacturing-like conditions took under 18 months, which for us is actually slow. With Canopy and the acceleration it gives us, we're confident the next one will take half that, if not less.
Satnam Surae and twig’s CEO Russ Tucker at Nebius AI Discovery Awards. Credits: twig
TIL: How far does Canopy's scope extend: is it just sunscreen and everyday chemicals, or more than that?
Surae: Much more. We built Canopy to be as multi-purpose and generalizable as possible for biomanufacturing. Beyond the sunscreen molecules, there are related ones that work as antioxidants, others that could go into flavors and fragrances, and others still that could be pharmaceutical ingredients, even drug molecules. Really, across the board.
'It's more engineering than science'
TIL: What's the failure mode you're most worried about?
Surae: It's more engineering than science. We can prove the science works in the lab; the real risk is getting something that works in the lab to work at 100,000-liter scale as an engineering process. Iteration is slower there, costs are larger, and it comes down to techno-economics more than anything technical or scientific. That's part of why Canopy factors in biomanufacturing-relevant data when it runs inference: it minimizes that risk as much as we can. Our mantra has always been to fail quickly and generate data even from failed experiments, because that data is still valuable.
TIL: And on compute: is this an inference-heavy workload?
Surae: There are two parts. Training is heavy, but critically, so is inference: we're running thousands of separate analyses per molecule, across thousands of molecules, in parallel. That gets extremely heavy. But partnering with someone like Nebius on inference, going forward, is something we're looking at to help maximize that potential.
TIL: You picked up second place in the Genomics category at the Nebius AI Discovery Awards in July 2026. What's your plan with the credits?
Surae: We’ll use the credits to accelerate our scaling plans for Canopy. We’re increasingly training across hundreds of millions of biological and experimental data points, so access to large-scale GPU infrastructure with the requisite availability is critical.
Disclosure
Surae spoke with The Infinite Loop at the Nebius AI Discovery Awards ceremony in London, where twig bio won second place in the Genomics category. The Infinite Loop is a Nebius editorial project. Editorial decisions are made independently.
