Latent Labs Brings AI-Powered Protein Design to the Browser with FoldFlow

In a move that could reshape how scientists and startups approach protein engineering, Latent Labs has launched a web-native AI model for protein design accessible directly from a browser, no code or compute cluster required.techcrunch.com

The product, called FoldFlow, brings large generative protein models out of the GPU farm and into the hands of biologists, researchers, and curious developers worldwide. Think of it as Figma meets Alpha Fold, a collaborative, visual interface for building novel proteins with structure-aware AI suggestions at every step.

A Playground for Molecular Creators

FoldFlow runs entirely in the browser using WebGPU. Users can drag and tweak protein sequences and get real-time structural predictions, powered by a distilled version of Latent Labs' flagship generative model. The interface allows point-and-click modifications, visual confirmations of protein folds, and live feedback on biochemical plausibility.

According to Latent Labs co-founder and CEO Kyle Michel, “We want protein design to feel like design, not simulation. It should be fast, visual, and intuitive.”

At its core, FoldFlow is a fine-tuned diffusion model trained on hundreds of thousands of known protein structures. It operates like an architectural assistant given a partial or rough sequence; it suggests viable 3D configurations, fixes unstable loops, and proposes mutations that optimize for solubility or function.

Early testers have used it to prototype enzymes, binding proteins, and structural scaffolds for therapeutics, all without writing a single line of code.

What’s notable is not just the performance, it’s the packaging. By offloading inference to lightweight models that run in the browser (with optional cloud acceleration), Latent Labs eliminates one of the biggest barriers to entry: GPU access. This levels the field for small labs, indie biotech startups, and even students exploring protein modeling for the first time.

FoldFlow also integrates tools like visualization, mutational scanning, and export options for wet-lab synthesis. It’s designed to work end-to-end—idea to candidate protein within one interface.

From Research to Product, Fast
The company, backed by a mix of deep tech and biotech investors, says it sees applications across drug discovery, synthetic biology, and material science. "We’ve seen protein design move from PhDs in structural biology to AI researchers. Now, it’s about putting it in the browser so a designer in a biofoundry can collaborate on a molecule like they would on a slide deck," said co-founder Chris Hallacy.

While the field of generative biology is crowded with players like Generate Biomedicines, Cradle, and Profluent advancing their own AI models, Latent Labs is betting on usability and speed over brute-force prediction. Their strategy: reduce friction, expand access, and let users build with proteins like they're prototyping digital products.

Available Now, Built for the Future:
FoldFlow is live and free to use at Latent Labs’ website, with pro tiers for team collaboration and cloud inference acceleration coming later this year.
In an industry often gated by complexity and compute costs, Latent Labs is making a clear statement: protein design doesn’t have to be locked in the lab. It can live in the browser.