24 September 2026

^{Enveda is a biotechnology company that aims to learn from life's chemistry to create better medicines faster (image: Enveda)} 

 

Put a plant extract through a mass spectrometer and it will produce a forest of chemical signals. Each offers clues about a molecule in the sample. Many of those molecules have no familiar name, and finding one that might make a useful medicine is slow work. First, scientists must identify it. Then they must discover what it does. Finally, they must work out whether it can become a drug.

Enveda, a biotechnology company based in Boulder, Colorado, is trying to speed up that search. Its AI drug discovery platform helps interpret the signals left by molecules produced by plants, microbes and human bodies. Chemists and biologists then investigate promising finds and turn them into candidates for testing.

On 23 September, Enveda announced a $311 million Series E financing, bringing the total it says it has raised since its founding to more than $845 million. The money will support further trials of three experimental medicines and expand the automated laboratory and AI platform behind them.

The question for Enveda now extends beyond whether AI can find interesting chemistry. Some of its discoveries have reached people. Can any of them become a useful treatment?

 

Reading what nature has made

 

Enveda was founded in 2019 by Viswa Colluru, who previously worked at AI drug discovery company Recursion. Enveda says he began building the business with $55,000 of his own savings. Its scientific co-founder, Pieter Dorrestein, is a researcher whose work has helped develop ways to study the molecules made by living organisms.

Their starting point is a longstanding problem in drug research. Natural compounds have inspired valuable medicines, but a plant or microbial sample can contain thousands of molecules. Even if the sample shows a useful effect in an experiment, identifying the compound responsible can be painstaking.

Enveda’s platform combines mass spectrometry, AI and laboratory testing. A mass spectrometer breaks molecules into fragments and measures the resulting pattern. Researchers can use that pattern to infer a molecule’s structure, though interpreting unfamiliar patterns is difficult. Enveda’s AI foundation model, PRISM, has been trained on large quantities of mass spectrometry data to help predict the structures of molecules that might otherwise remain unidentified.

The prediction gives scientists a lead, rather than a finished answer. Laboratory work must check the chemistry and establish whether a molecule affects the biology of a disease. Medicinal chemists must then improve promising starting points so they can be made reliably, reach the right part of the body and have a reasonable chance of being safe.

That division of labour explains what is distinctive about Enveda. Some AI drug discovery companies use models to design new molecules for a chosen target. Enveda concentrates on making existing natural chemistry searchable at scale, before its scientists refine selected molecules into potential medicines. It is a bet that living things have produced many useful chemical ideas that conventional research has struggled to find.

 

Enveda’s first drug candidates reach clinical trials

 

Enveda’s most advanced candidate is ENV-294, an experimental once-daily pill being investigated for atopic dermatitis, or eczema, and asthma. It has moved well beyond a computer prediction: the company has tested it in people and begun Phase 2 studies.

In a Phase 1b study involving nine adults with moderate-to-severe eczema, Enveda reported that participants’ eczema severity scores fell by an average of 85% by day 42. They took ENV-294 for 28 days, followed by two weeks of observation.

That result is intriguing, but the study was small and open-label: everyone knew they were receiving the drug. It had no placebo group. It cannot establish how much of the improvement was caused by ENV-294, whether a benefit will last or how the drug’s safety profile will look in a larger population. The ongoing studies are designed to provide stronger evidence.

Two other candidates show the range of Enveda’s ambitions. ENV-308 was inspired by Lac-Phe, a molecule the body releases during intense exercise. Enveda is developing an experimental pill for metabolic conditions, including the possibility of helping people maintain weight loss after stopping GLP-1 medicines. An initial study in 88 healthy volunteers provided information about safety, tolerability and biological activity. It did not show that the drug helps people keep weight off.

ENV-6946, an oral candidate for inflammatory bowel disease, entered Phase 1 testing in late 2025. It is intended to act on several inflammatory pathways, but whether it benefits patients remains unknown.

Together, the three candidates demonstrate that Enveda can move discoveries from its platform into human testing. They do not yet establish that its approach produces approved medicines, or that it does so faster or more cheaply than other methods.

 

Building the team for the next stage

 

Discovering candidates takes expertise in AI, analytical chemistry, biology and medicinal chemistry. Running larger trials calls for clinicians, manufacturing specialists and people experienced in drug development.

Enveda has been adding that experience alongside its discovery team. Its September financing is intended to move ENV-294 and ENV-308 into later-stage studies, advance ENV-6946 and bring further candidates into clinical testing. It is also funding more work on PRISM and the laboratory that supplies its data. The investment will test whether a platform built to find molecules can support a sustained pipeline of medicines.

Enveda is not alone in this space: several companies are taking an AI-native approach to drug discovery, although the term covers different kinds of work. Iambic Therapeutics uses AI to help design and optimise molecules. Its experimental cancer drug IAM1363 reached human trials in roughly two years, and the company has reported early tumour responses in some patients with HER2-driven cancers. Larger studies are needed to establish its effectiveness and safety.

Stanford researchers are exploring another route with a Virtual Biotech that can deploy thousands of specialised AI agents to analyse research and propose drug development strategies. It proposed an antibody–drug conjugate approach for a lung cancer target called B7-H3. The Stanford system did not make or test a drug of its own; laboratory experiments remain the test of its ideas.

These efforts show why claims about “AI-discovered drugs” need a closer look. AI can help researchers find a molecule, design one or decide which hypothesis to pursue. Each achievement is different, and none removes the need for careful experiments and clinical trials.

Enveda has made an enormous collection of chemical signals easier to read and taken three resulting candidates into human studies. For ENV-294, the next chapter is already underway. The answer will come from patients in larger trials, rather than from another promising signal in a machine.