AI-native CRO

A CRO that returns the experiment, not just the result.

Erid instruments wet labs so every study comes back as continuous, machine-readable physical data: high-resolution video, thermal imaging, and sensor telemetry, time-aligned to the endpoint you ordered. Delivered through one API.

Built for teams training world models of biology and chemistry.

video
1920×1080 · 30 fps
thermal
radiometric °C
telemetry
signal vs time
0–90 min

A spreadsheet is a lossy recording of a physical process.

An assay is hours of physical behaviour: a growth curve bending, heat moving through a plate, an optical signal drifting, a reagent added four seconds late. A traditional CRO compresses all of it into one number per well and discards the rest. That's the right output for a go/no-go decision. It's the wrong training signal for a model meant to learn how biology and chemistry behave.

Every study Erid runs is recorded end to end. You get the number. You also get what produced it.

What a CRO returns

wellvalueunit
A112.4nM
A211.9nM
A313.1nM
3 numbers ~40 bytes

What Erid returns

video

1080p · 30 fps

~9.2 GB

thermal

calibrated °C

~3.1 GB

environment

T, RH, CO₂

~120 MB

instrument telemetry

raw signal

~1.8 GB

protocol trace

timestamped

~2 MB

endpoint

12.4 nM

~240 B
~14 GB one manifest
Six streams, one clock

Every artifact from a run carries the same synchronised timestamp, so a frame of video, a thermal reading, and a step in the protocol can be joined without guessing.

StreamStatus

Visual

Continuous high-resolution video of the bench, plate, or reaction vessel

Live

Thermal

Radiometric IR imaging, per-pixel temperature

Pilot

Environmental

Ambient temperature, humidity, CO₂, light, vibration

Live

Instrument telemetry

Raw instrument output, pre-processing

Live

Protocol trace

Every step and operator action, timestamped, deviations flagged

Live

Endpoint

The assay result you ordered

Live

Read the full data specification →

01 From API call to training data
01

Specify

Send a study as a structured request: the assay, the conditions, and the capture spec for which streams you want, at what resolution and rate.

02

Run

The study executes in a partnered wet lab retrofitted with Erid sensor pods. Established facilities, existing accreditation, existing technicians.

03

Capture

An edge-compute pod time-syncs every sensor to the protocol trace, compresses on site, and uploads continuously. You see data while the study is running, not weeks after it closes.

04

Deliver

Streams land in your bucket or come down through the API. One manifest per run ties every artifact to the protocol step that produced it.

The retrofit

We don't build labs. We instrument them.

Building an automated lab in the US costs tens of millions and takes years. Erid takes established wet labs in India's Genome Valley (real facilities, real technicians, real throughput) and adds what they're missing: cameras, thermal, environmental sensing, and an edge-compute pod that synchronises and streams everything off site.

You get automated-lab-grade observability without automated-lab capital, at a fraction of the cost of building it yourself.

camiredgeenv
Who this is for

Built for teams training models that have to touch the physical world.

Δ

Foundation models for biology and chemistry

Teams whose training data currently stops at published endpoints and simulation.

Autonomous lab companies

Teams that need physical grounding at volumes their own facility can't reach yet.

Pharma AI groups

Teams with in-house models and an existing CRO spend returning nothing reusable.

Vision
Now

An asset-light data layer on top of partnered labs, so the streams exist today.

Next

Erid-operated facilities, purpose-instrumented rather than retrofitted.

Then

Fully autonomous robotic wet labs, the foundational CRO infrastructure for the AI era.

Tell us what you're training.

Send us the model and the physical behaviour it's missing. We'll design the capture spec and quote a pilot run.