An automated experiment loop for research labs
Discovery Loop connects to your instruments, reads their output, infers what to test next, and queues the next experiment. The cycle runs without a researcher between steps.
Four stages in one loop
Each stage passes its output to the next. There is no export, review pause, or batch scheduling step. The loop moves at instrument speed.
Stage 01
Ingest
Instrument adapters collect raw output from spectroscopy, chromatography, plate reader, and sequencing hardware as soon as each run ends. Vendor specific schemas are normalized on arrival.
Stage 02
Analyze
A structured analysis model reads normalized data, extracts key signals, and updates the hypothesis tree. It flags anomalies, assigns confidence intervals, and marks branches open or closed.
Stage 03
Choose
The decision engine picks the next experiment from the open hypothesis space through an information gain policy. It weighs cost, instrument availability, and uncertainty when ranking candidates.
Stage 04
Send
The chosen experiment becomes a protocol and moves to the instrument queue. Scheduling accounts for lead times, batch sizes, and reagent limits set in your lab configuration.
Instrument data, ready to use
Instrument vendors use different file formats. Discovery Loop reads them natively, removing the need for middleware or a custom ETL pipeline.
Connected from day one
Add your API credentials and Discovery Loop starts reading files from the instruments already in your lab. Adapters support spectral peak tables, chromatography retention data, plate reader matrices, and flow cytometry cell counts.
An unsupported instrument can be added by your team in under a day with the open adapter SDK. Every adapter uses the same normalized schema, leaving the loop logic unchanged.
The loop picks the next run
Information gain over intuition
The decision engine ranks candidate experiments by expected information gain per unit of cost. It favors the runs that reduce uncertainty fastest across the active hypothesis space, rather than those that validate a researcher's initial guess.
The policy can be configured. Limit candidates to available reagents, assign cost weights by instrument, or favor safety critical parameter ranges in materials work.
Hypothesis tree tracking
Each cycle adds evidence to a growing tree of conditional hypotheses. Ruled out branches are pruned, while new possibilities extend the tree.
Configurable policy
Set cost weights, instrument constraints, and exploration-exploitation balance through the lab config file or API.
Decision audit
Every recommendation is recorded alongside its supporting evidence, allowing you to trace a decision to the precise instrument run that redirected the loop.
Score = expected information gain / run cost
Fits your lab stack
Discovery Loop connects to your lab's data systems. No migration required.
Ready to run the loop in your lab?
Talk to the team about connecting your instruments and running your first automated cycle.