Making human microbiology computable
Most biological frontier models focus on what is encoded in human DNA. But our health is shaped by interactions between human and microbial biology, a layer that remains largely unmodelled. Outpost Bio is building AI at that interaction layer.
How the loop works
Our lab-in-the-loop platform combines high-throughput wet-lab assays with machine learning in a continuous feedback loop. Models learn from experimental results and guide the next round of testing.
Capture
Capture advanced multi-omics readouts to establish both the composition of the microbiome and its function, pre and post intervention.
Experiment
Experiments test how drugs, food fibers, and other compounds interact with the microbiome, at scale and under controlled conditions.
Train
Train Outpost frontier models on the experimental results and open access data, improving prediction performance with each new data point.
Design
Design improved experiments using model predictions, closing the loop and speeding up discovery with each cycle.
Want to work on this?
We are hiring across research, modeling, engineering and operations, in London and Boston.
See open roles
The mission
The microbiome shapes how we respond to drugs, food and therapeutics, and it is the largest unmodeled variable in human biology. We are building the AI layer for it.
The result we are aiming at: safer, more effective drugs, ingredients and interventions, designed with a fuller picture of the biology that drives outcomes.
The result we are aiming at: safer, more effective drugs, ingredients and interventions, designed with a fuller picture of the biology that drives outcomes.
What we are building
Atlas is a 539,308 data point pretraining corpus curated from public data, the largest dataset assembled for microbiome modeling. Waypoint is a family of open-source GPT-style causal language models from 6M to 170M parameters, trained on Atlas. Compass is an eight-task benchmark spanning biome classification, drug-microbiome interactions, drug degradation and infant gut development.
Behind them sits Expanse, our in vitro screening and ML validation service at LabCentral in Cambridge, MA, and the ingestion pipelines that harmonise multi-omic data across sites, sequencing technologies and protocols.
Behind them sits Expanse, our in vitro screening and ML validation service at LabCentral in Cambridge, MA, and the ingestion pipelines that harmonise multi-omic data across sites, sequencing technologies and protocols.
Our offices
A wet lab in Cambridge, Massachusetts and a team in London, with two days a week from home and two weeks a year from anywhere.
Boston
LabCentral, 700 Main Street
Cambridge
MA
02139
US
London
London
GB