Skip to content

Intelligence for engineers who build the real world.

SEIB is a foundation model for industrial engineering and Physical AI. It gives engineers the intelligence and the tools to design, automate, validate and improve complex industrial systems — faster. It runs on your network, including fully air-gapped.

Specification · Rev. APre-launch · pilot-partner phase
01Model class
Industrial foundation model
Continued pretraining on a verified engineering corpus
02Lifecycle span
Intent → physical action
Product, process, automation, factory, floor
03Engineering inputs
Requirements · CAD · BOM
Variants and tolerances read as structure, not text
04Automation output
PLC · robot · vision · tooling
Generation, conversion, and simulation
05Verification
Toolchain-gated
Checked against real compilers and runtimes first
06Factory context
Machines · cells · live data
Constraints and real-time signals over OPC UA
07Network
Cloud · on-prem · air-gapped
No dependency on the control network
08Safety posture
Proposes · never deploys
Hard-blocked from safety-related program units

You solve the hard problems · SEIB gives you the clarity, context and intelligence to move forward with confidence · no public API or price list yet.

The challenge

Engineering is complex.
Change is constant.
Time is critical.

Industrial engineering is fragmented and heavily dependent on scarce expert knowledge. In high-mix, NPI-driven factories, products, variants, processes and equipment change constantly — but engineering and commissioning stay too manual and too slow to keep pace.

Too many systems

Data and tools are siloed across the engineering stack. Requirements, CAD, process plans and controls each live in their own format, and the work of reconciling them falls to people.

Change happens faster than engineering

New products, variants and processes emerge constantly. Engineering and commissioning stay manual, so the factory is always catching up to a product that has already moved on.

Too dependent on scarce expertise

Critical knowledge lives in people's heads. That creates risk when they are unavailable and a bottleneck whenever more than one line needs them at once.

The cost of delay is too high

NPI and line changes take too long and impact business performance directly. Every week of ramp-up is a week the line is not producing what it was built to produce.

The SEIB model

One foundation intelligence across the full engineering lifecycle.

SEIB connects product intent to physical outcomes — understanding, reasoning and generating what engineers need at every step. Every draft clears a real toolchain on its own before it becomes something you review.

Sequence · one request, one scanCycle complete · held for review
  1. 01

    Product intent

    Requirements, CAD, BOM, variants and tolerances — read as structure, not as documents to search.

  2. 02

    Process design

    Workflows, operations, quality checks and cycle-time targets, derived from what the product actually is.

  3. 03

    Automation engineering

    PLC, robot, vision and tooling code, generated in the target language and simulated before release.

  4. 04

    Factory context

    Machines, cells, constraints and real-time data — the plant as it is today, not as the drawing says.

  5. 05

    Physical action

    Commission, operate, adapt and improve — with every change held for a human to approve.

A person approves every change. Approved changes go through your own CI to the machine. SEIB is hard-blocked from writing to safety-related program units, and from deploying anything itself.

Automation engineering · request → generated diffMixer_Line3.st
Engineer’s request

“Add a 5-second startup delay and an E-stop interlock to the mixer motor on Line 3.”

Generated · Structured Text
IF NOT E_Stop AND StartCmd THEN
  StartTimer(IN := TRUE, PT := T#5s);
  IF StartTimer.Q THEN
    MotorRun := TRUE;
  END_IF;
ELSE
  StartTimer(IN := FALSE);
  MotorRun := FALSE;
END_IF;
Compiled clean · 0 errorsSimulated 1,200 cyclesDiff ready for review
Why it's different

A model, not a wrapper around one.

Most industrial copilots are a chat window in front of a general-purpose model. SEIB is trained on the thing it's supposed to be good at.

Generic AI layer
SEIB
The model
A general-purpose LLM pointed at your documentation.
Trained from the ground up on verified industrial engineering work.
Scope
One task at one step, disconnected from everything around it.
One intelligence from product intent through to physical action.
Verification
Trusts its first answer. You find the error in commissioning.
Every draft checked against a real toolchain before you see it.
Deployment
Cloud-hosted, with local as an afterthought — if offered at all.
Built to run fully air-gapped from day one, on your own hardware.
What it knows
Your manuals and support tickets, retrieved on request.
Your CAD, your cells, your constraints, your own plant's conventions.
Deployment & safety

Nothing here touches your plant on its own.

The question every plant asks first, answered before you have to ask it.

Your data stays on your site

SEIB runs on your network — cloud, on-prem, or fully air-gapped. Engineering project files never have to leave the building for the model to read them.

A human commits every change

SEIB proposes a reviewable diff. It never writes to a live machine and is hard-blocked from safety-related program units, regardless of deployment mode.

Built for zone-and-conduit thinking

Architected in alignment with IEC 61508 lifecycle practice and IEC 62443 network segmentation — designed to sit cleanly inside an OT security model your team already runs.

SEIB is pre-certification. “Aligned with” and “architected for” describe design intent, not a claim of formal certification against IEC 61508 or IEC 62443 by an accredited body.

Foundation intelligence for industrial engineering

Mission

Foundation intelligence for industrial engineering.

SEIB helps manufacturers understand what needs to change in the factory when the product changes — across engineering, commissioning, and the Physical AI that runs on the floor.

Vision

Factories that understand what to build, and how to adapt themselves.

Industrial systems that can reason, learn and evolve — just like the best engineers. One layer of intelligence the plant can think with, rather than a dozen tools it has to be translated between.

Built for industries that build the world
  • Automotive
  • Electronics
  • Industrial Equipment
  • Consumer Products
  • Aerospace & Defense
  • Energy
Questions

Questions we get before the first call.

The short answers. If yours isn't here, the pilot form below reaches a person.

01What is SEIB?

SEIB is a foundation model for industrial engineering and Physical AI. It gives engineers the intelligence and tools to design, automate, validate and improve complex industrial systems. SEIB is a model trained on industrial engineering work, not a chat window in front of a general-purpose model.

02What does SEIB cover across the engineering lifecycle?

SEIB spans five connected stages: product intent (requirements, CAD, BOM, variants, tolerances), process design (workflows, operations, quality, cycle-time targets), automation engineering (control, robot, vision, tooling, simulation), factory context (machines, cells, constraints, real-time data), and physical action (commission, operate, adapt, improve).

03What kind of engineering work can SEIB generate?

SEIB generates and converts automation artefacts, including control code in the IEC 61131-3 languages, robot and vision routines, and tooling and simulation configuration. Each output is produced in the target format of the toolchain the plant already runs.

04Can SEIB run air-gapped or on-premises?

Yes. SEIB deploys cloud, on-premises, or fully air-gapped, and was built for air-gapped operation from day one rather than as an afterthought. SEIB has no dependency on the control network, so engineering project files never have to leave the building for the model to read them.

05Does SEIB write directly to machines on the factory floor?

No. SEIB proposes a change as a reviewable diff and a person approves it. Approved changes go through the customer's own CI to the machine. SEIB is hard-blocked from writing to safety-related program units and from deploying anything itself, in every deployment mode.

06How is SEIB different from ChatGPT or a generic coding copilot?

Industrial engineering formats are low-resource for general-purpose models, so those models produce output that reads correctly and fails against a real toolchain. SEIB is trained on verified industrial engineering data and reinforced against real compilers, runtimes and simulators, and every draft clears that toolchain before an engineer sees it.

07Which toolchains and protocols does SEIB work with?

SEIB sits beside the engineering tools a plant already runs rather than replacing them, across CAD, process planning and controls environments. SEIB performs protocol-aware reasoning and tag resolution over OPC UA and Modbus to read live factory context.

08Which industries is SEIB built for?

SEIB is built for industries that build the world: automotive, electronics, industrial equipment, consumer products, aerospace and defense, and energy. The common thread is high-mix production where products, variants and processes change faster than engineering can keep up.

09How do I get access to SEIB?

SEIB is in a pre-launch, pilot-partner phase, with no public API or price list yet. Request pilot access through the form on seib.ai or email hello@seib.ai. It is not a waitlist — every request is read by a person.

Pilot access

Let’s build the future together.

Not a waitlist. Tell us the problem and we'll tell you honestly whether SEIB is ready for it.

Pilot request · confidential0 of 5 complete