Autonomous R&D Infrastructure

Make your lab autonomous.

Nornec deploys physics-aware agents that monitor, interpret and steer experiments around the clock — on the instruments you already own.

Existing hardware. No rip-and-replace. Human authority over high-consequence actions. Every decision traceable.

Your scientists should decide what matters — not babysit instruments.
TENSOR / Lab Control Loop ● LIVE
Measure
Oscilloscope · spectrometer · PLC · flow system
Understand
Physics-aware state · regime classifier · drift gate
Decide
Policy check · next run · human escalation
Act & Learn
Setpoint update · execute · log · feedback
runs: 8,421 logged: 127,934 mode: autonomous
Channels online
< 1 s
Control-loop latency
ARMED
Safety state
24/7
Runs completed

Built on real instruments. Running real experiments.

Names and protocols we actually connect to.

Nornec integrates with hardware already on your bench. No rip-and-replace. No phantom roadmap. Every integration below is either connected in a live deployment, validated in Lab 242, or supported through a standard interface we can document.

Rohde & Schwarz
VALIDATED
Rigol
CONNECTED
BK Precision
CONNECTED
SRS / RGA
VALIDATED
Timepix3
CONNECTED
PMT / MCA
CONNECTED
LibreVNA
VALIDATED
PLC / Modbus
SUPPORTED
OPC-UA
SUPPORTED
SCADA feeds
SUPPORTED
Local GPU cluster
CONNECTED
Experiment database
SUPPORTED
Connected — live in a deployed loop Validated — tested end-to-end in Lab 242 Supported — standard interface available

Connect what you already own

One auditable control loop for every instrument.

Connect oscilloscopes, spectrometers, power supplies, PLCs, mass-flow systems, gas analyzers, radiation detectors and process sensors into one auditable control loop.

Oscilloscopes
Spectrometers
Power supplies
PLCs
Mass-flow systems
Gas analyzers
Radiation detectors
Process sensors
Deployed through standard interfaces Tested through standard interfaces Supported through standard interfaces

Evidence of a real run

It rejects dangerous actions. Then starts the next run.

A timestamped excerpt from a live TENSOR session. The system enforces hard safety envelopes, logs every decision, and only escalates to a human when the consequence exceeds policy.

14:23:01.847 [MEASURE] 47 channels online · RTO2014, BK Power Supply, Rigol DG922, RGA120, LibreVNA
14:23:04.112 [UNDERSTAND] Regime classifier: stable plasma · H₂ flow 8.2 mL/min · pressure within envelope
14:23:04.903 [DECIDE] Proposed: increase pulse voltage +4.2 % for run #0842
14:23:05.021 [POLICY] REJECTED — proposed setpoint exceeds hard over-voltage envelope (420 V limit)
14:23:05.118 [DECIDE] Fallback selected: +2.1 % sweep within envelope · no human escalation required
14:23:06.205 [ACT] Execute run #0842 · all channels logging · safety state ARMED
14:23:12.671 [LEARN] Run #0841 archived · residuals within 3 % · update instrument drift model
14:23:13.004 [STATUS] Next run queued · operator notification: low priority · decision provenance: 7f3a9d2
8,421 autonomous runs
134 anomalies surfaced
unattended operation
full decision provenance
Representative TENSOR session · live data varies by deployment

Loop architecture

Measure → Understand → Decide → Act & Learn.

Four stages you can explain to a board. Under the hood, seven auditable services make it real.

01 — Measure

Ingest & Validate

Multi-channel ingestion with provenance tagging. Stale, missing or out-of-envelope channels are rejected before compute.

02 — Understand

Compute & Hypothesize

Physics-aware state estimation, regime classification, and testable hypotheses — checked against dimensional and conservation constraints.

03 — Decide

Policy & Orchestrate

Role-based policies convert hypotheses into run plans. High-consequence actions queue for human approval; routine sweeps proceed.

04 — Act & Learn

Execute & Analyze

Run the campaign, archive every setpoint and decision, then feed residuals and insights back into the model for the next loop.

01 Ingest
02 Validate
03 Compute
04 Hypothesize
05 Decide
06 Execute
07 Analyze & learn

State estimation

Physics-aware state estimation.

Live telemetry is fused with instrument models, experimental history and known physical constraints to estimate system state, detect drift and identify genuine regime transitions.

  • Dimensional consistency checks
  • Conservation residual monitoring
  • Instrument drift models
  • Regime-transition falsification gate
STATE ESTIMATE // RUN #0841
Plasma impedance modelmatched H₂ production rate8.2 mL/min Optical emission ratiowithin regime Pressure envelopeOK Voltage envelopeOK Drift correctionapplied
regime: stable pulsed plasma · next action: proceed to run #0842

Business outcome

More experiments. Fewer wasted campaigns. No overnight babysitting.

Run continuously

Keep campaigns alive after hours, weekends and holidays without leaving a scientist at the bench.

Stop bad runs early

Detect drift, anomalies and unsafe envelopes in seconds — not after the next morning’s data review.

Preserve every decision

Full provenance for setpoints, approvals, model versions and escalations. Audit-ready by default.

Compound experimental knowledge

Residuals and outcomes feed back into the model so the system learns from every run.

Keep scientists in control

Autonomous where safe, escalated where it matters. Human authority over high-consequence actions.

Not another dashboard

Conventional lab software vs Nornec Lab OS.

Conventional lab software Nornec Lab OS
Data handling Collects and visualizes Ingests, validates, and reasons from physics
Next action Scientist decides manually System proposes, policy approves, human escalates
Safety Static alarms Hard operating envelopes with automatic rollback
Provenance Files scattered across tools Every decision logged with model version and policy
Deployment Cloud-only or siloed On-prem or private-cloud, no raw data leakage
Learning Each campaign starts from zero Residuals and insights compound across runs

Integrations

We do not ask you to replace the lab. We make the lab you already own autonomous.

Scientific instrumentation

Oscilloscopes, spectrometers, VNAs, particle detectors, radiation counters, cameras and custom front-ends.

Experiment control

Power supplies, arbitrary waveform generators, motion stages, gas and liquid mass-flow controllers, valves and relays.

Industrial systems

PLCs, DCS interfaces, SCADA feeds, safety interlocks, environmental monitors and building management systems.

Data and compute

Time-series databases, object stores, local GPU clusters, private cloud and existing notebook or analysis environments.

Safety and governance

The agent recommends. The policy decides. Physics has the final word.

On-prem / private-cloud

Raw instrument data stays inside your network. No uncontrolled external model access.

Role-based policies

Who can approve what is configured by your team, not hard-coded by us.

Hard operating envelopes

Voltage, pressure, temperature, flow and dose limits enforced below the agent layer.

Human approval

High-consequence actions queue for explicit human sign-off before execution.

Automatic rollback

Unsafe drift triggers fallback setpoints and orderly shutdown sequences.

Model versioning

Every deployed model, policy and threshold is versioned and reproducible.

Decision provenance

Each action carries a traceable hash linking model, policy, input state and operator.

No uncontrolled external model access

Models run on your infrastructure. External APIs are opt-in, scoped and logged.

“The agent may recommend an action. The policy determines whether it is allowed. Physics determines whether it is possible.”

How to buy

Productized autonomy. Fixed scopes. No slideware.

1. Autonomy Fit Call

30 minutes · No fee

We review your instrument park, experiment type and safety constraints to tell you where autonomy is ready and where it needs guardrails.

2. 48-Hour Lab Autonomy Audit

From USD 12,500

Instrument topology map, agent-readiness scorecard, safety envelope review and a 90-day roadmap to closed-loop operation.

3. Closed-Loop Pilot

USD 75,000–150,000

One autonomous loop deployed on your hardware: ingest, state estimation, policy, execution and logging — with human escalation.

4. Lab OS Deployment

USD 250,000+

Full multi-loop deployment, model governance, runbook, training and continuous improvement across your lab.

Energy systems

Still grounded in process physics.

Lab autonomy sits on top of the same first-principles discipline that powers our energy and process work. We don't sell generic software. We deploy autonomy into real laboratories.

Nuclear Infrastructure Technical Due Diligence

SMR validation, data-center-nuclear interconnect DD, advanced reactor design review.

Frontier Energy R&D Advisory

Fusion program technical advisory, plasma diagnostics architecture, IP/patent strategy.

Technical Due Diligence — Complex Low-Carbon Infrastructure

Bespoke DD on multi-vendor, multi-technology low-carbon projects.

Nornec vs Aidevo

Who builds what.

Nornec owns the scientific judgment, experimental architecture and deployment. Aidevo provides the agent infrastructure, admissibility kernels and engineering engines underneath.

Nornec decides what the lab should do and validates that it is physically sound. Aidevo supplies the runtime, models and platform components that make autonomy repeatable and scalable.

Visit Aidevo →

“If a claim sounds too good, that's the one to send.”

Tell us what you're evaluating. We'll tell you what the physics permits — directly, and in writing.

szg@nornec.com →

Deploy your first autonomous experiment.

Start with a 48-hour Lab Autonomy Audit. We'll map where your lab is ready to run itself — and where humans must stay in charge.

START THE 48-HOUR AUDIT →