The companies we back, and why we're excited about each of them.
Pills where there used to be injections: AI-designed peptide drugs, already partnered across big pharma.
Most of biology's most valuable drug targets have been out of reach for pills. Unnatural Products designs macrocyclic peptides, a class of molecules large enough to grip those targets and engineered to be taken by mouth, using AI-guided design and parallel synthesis at a scale traditional chemistry can't match.
AI built the way the brain keeps time, learning from far less data and compute.
Today's AI gets better by getting bigger. Cognitive Scientific is betting on a different route: models built on how the brain actually represents time and memory, an approach its founders, Marc Howard of Boston University and Per Sederberg, have developed through more than a decade of peer-reviewed research.
Placebo-controlled trials at the scale of the supplement aisle.
Billions of dollars of wellness products are sold every year with almost no evidence behind them. Radicle Science runs large, decentralized, placebo-controlled trials that give those products real proof, and has enrolled nearly 100,000 Americans in its trials since 2021.
Earbuds that read your brain while you sleep.
NextSense makes Smartbuds, wireless earbuds with built-in EEG that measure brain activity through the night and play audio tuned to deepen sleep. It's the kind of measurement that used to need a sleep lab, now in something people will actually wear.
Turning a whole genome and a lifetime of records into a plan a doctor can act on.
Personalized medicine has a paperwork problem: a single patient can arrive with a genome and hundreds of pages of history that no physician has time to read. Profile Health is a clinical operating system that synthesizes all of it, genetics, records and clinical data, into something a doctor can use in minutes.
An autonomous machine-learning engineer: plain English in, deployed model out.
Every company sitting on its own data wants models trained on it, and almost none can hire the machine-learning engineers to build them. Impulse turns a plain-English request and a dataset into a production model, with no data-science team required.
A lab in every toilet, starting with cortisol and testosterone.
Your body's chemistry changes every day, and medicine checks it a few times a year. Trace Health is building a passive sensor for the toilet that reads urine every time it's used: no test strips, no apps to remember, no change in behavior.
Enzymes engineered to change how the body processes food.
Metabologic designs enzymes that work on what we eat, aiming to improve how the body handles food for better health, longevity and outcomes.
The retrieval database AI apps reach for first, becoming the context layer for agents.
When a developer needs a model to find the right information, Chroma is usually what they install first. The open-source database is downloaded more than 15 million times a month and comes built into LangChain and LlamaIndex.
