Beat 01
Shadow
Observes and predicts. No write-back exists. Accuracy is scored against your own dispositions.
FAQ
Direct answers about trust, integration and pricing
The questions fabs actually ask in the first three meetings, answered without hedging.
- Autonomy
- Integration
- Security
- Commercial
Trust and autonomy
How much control does it take, and when?
Through three gates. Shadow mode observes and predicts with no write-back; advisory mode recommends recipe, transfer and binning moves that a process-integration or yield engineer approves; graduated autonomy releases low-risk deposition, anneal and transfer control once measured accuracy and twin validation clear the bar. High-impact decisions stay human-in-the-loop.
Reducing false calls is the wedge, not a side effect. The mura-and-defect agent is trained on true-versus-false-call labels from your own inspection history, so it separates genuine large-area non-uniformity, sub-pixel opens and shorts, particles and stains from nuisance signals, and every call is traceable to the evidence behind it.
Per-tenant isolation with recipes, panel designs and defect images scoped to your tenant, encryption in transit and at rest, SSO/RBAC, an immutable yield/quality-grade audit log and an on-prem or air-gapped option. Fleet learning shares model improvements, never your recipes.
Autonomy
The three gates, restated plainly
Beat 02
Advisory
Recommends recipe, transfer and bin moves. An engineer approves or overrides, and every override becomes a training label.
Beat 03
Graduated autonomy
Executes low-risk actions after sustained accuracy and twin validation. Released per agent, not all at once.
Beat 04
Always human
High-impact recipe, scrap and bin decisions stay human-in-the-loop by policy, at every tier.
# closed loop, one panel
perceive substrate → tft → emitter stack → sub-pixel
plan deposition · anneal · transfer · test/bin
act recipe write-back (SECS-GEM)
sense mura ΔL* · sub-pixel opens · particles
optimise yield · ramp · scrap
log immutable audit entry ✓ signed
retrain engineer correction → model
Integration
Fitting into a fab that already runs
Deposition, photolithography, encapsulation and anneal tools, Mura/AOI inspection and metrology, array and cell test, mass-transfer and repair stations, robotic handling and MES — vendor-neutral via SECS-GEM/HSMS where the tool supports it, with REST and OPC-UA bridges elsewhere.
The fab edge runtime keeps running without the cloud, agent steps are idempotent, and every control path has a fail-safe stop that returns tools and robotic handling to a known-good state. Degradation drops autonomy back to advisory rather than acting on a weak signal.
A design-partner pilot on one wedge workflow: instrument in shadow mode, measure the yield, false-call and ramp baseline, move to assist mode, and convert on a single agreed success metric tied to the pricing metric.
No. Emiteon reads from the inspection and metrology systems you already run and makes their output more useful by separating real excursions from nuisance and attaching the evidence behind every retained call.
Evaluation
What a first engagement looks like
Six steps, roughly a quarter, with a decision at the end.
Step 1 of 6 · scope
Scope
One workflow, one metric, named owner.
Step 2 of 6 · review
Security review
Architecture documentation to IT/OT and IP.
Step 3 of 6 · connect
Connect
Read-only connectors on the wedge tool stack.
Step 4 of 6 · baseline
Baseline
Measured from your own production data.
Step 5 of 6 · shadow
Shadow
Predictions scored, no write-back.
Step 6 of 6 · decide
Decide
Convert on the metric, or do not.
Deployment and data
Where things run and who owns what
Shadow-mode scoring typically produces a defensible read within eight weeks of connector commissioning, assuming historical dispositions are available for baselining.
Yes, including fully air-gapped. In that mode the control plane is local, model updates arrive as signed bundles through your own media process and no telemetry leaves the plant.
It is yours. Audit records and models trained on your dispositions are exportable in signed form, and your tenant data is deleted on request under the terms of the agreement.
Quick answers
One-line versions
Does it need the cloud?
No. On-prem and fully air-gapped deployment are supported.
Does it replace engineers?
No. It removes the nuisance work and captures their corrections.
Does it work with our tools?
Vendor-neutral via SECS-GEM/HSMS, with OPC-UA and REST bridges.
Is the pilot free?
No. Paid and time-boxed, with one agreed success metric.
Do you charge per user?
No. Pricing is per tool or line, per fab, or custom.
Can we stop?
Yes, and your data and audit records export in signed form.
Commercial and company
Pricing, ownership and where we are
Models trained on your dispositions are scoped to your tenant. Fleet learning shares generalised model improvements only — never your recipes, panel designs or defect imagery.
Yes, through the Enterprise tier, with central autonomy and approval policy, per-site exceptions and one consolidated audit trail.
Per tool or inspection line for the Line tier at $16,000 per month, per fab for the Fab tier at $110,000 per month, and custom for Enterprise, typically landing between $700k and $8M ACV.
Emiteon is at design-partner stage. Company operating status, customer references and outcome metrics described on this site are aspirational or placeholder until independently verified, and we label them that way deliberately.
Still unclear?
Ask the awkward version of the question
The questions that make a vendor uncomfortable are usually the important ones. Ask about escape rates, about what happens when the model is wrong, about our stage as a company. We will answer plainly.
- What is your escape rate when you suppress calls?
- What happens when the network drops mid-write?
- Who is liable if a recipe change scraps a lot?
- How many production lines are you actually live on?
# closed loop, one panel
perceive substrate → tft → emitter stack → sub-pixel
plan deposition · anneal · transfer · test/bin
act recipe write-back (SECS-GEM)
sense mura ΔL* · sub-pixel opens · particles
optimise yield · ramp · scrap
log immutable audit entry ✓ signed
retrain engineer correction → model
Voices from the line
What the fab floor tells us
“We do not lose panels because nobody is watching. We lose them because the signal that mattered was buried under a thousand nuisance calls.”
“Ramp is the whole game. If a new product takes two quarters of manual tuning, that is two quarters of margin we never get back.”
“Transfer yield is our ceiling on microLED. Every dead emitter is a repair cycle or a scrapped backplane.”
Quotes are illustrative composites drawn from discovery interviews with process-integration, yield and quality engineers. Named references are [PLACEHOLDER] pending design-partner consent.
The economics
What the loop is measured on
Emiteon is priced and evaluated on the numbers a fab already reports. These are design-partner targets for the first twelve months of deployment.
$26BDisplay-fab automation, process control, inspection, test and fab software market
~14%Annual growth in the segments Emiteon plays in
135%Net revenue retention target from land-and-expand
99.9%Uptime target for the fab edge runtime
Figures marked as targets are design-partner objectives, not audited results. Company operating status, customers and outcomes are [ASPIRATIONAL] until independently verified.
FAQ
Questions fabs ask first
Through three gates. Shadow mode observes and predicts with no write-back; advisory mode recommends recipe, transfer and binning moves that a process-integration or yield engineer approves; graduated autonomy releases low-risk deposition, anneal and transfer control once measured accuracy and twin validation clear the bar. High-impact decisions stay human-in-the-loop.
Reducing false calls is the wedge, not a side effect. The mura-and-defect agent is trained on true-versus-false-call labels from your own inspection history, so it separates genuine large-area non-uniformity, sub-pixel opens and shorts, particles and stains from nuisance signals, and every call is traceable to the evidence behind it.
Per-tenant isolation with recipes, panel designs and defect images scoped to your tenant, encryption in transit and at rest, SSO/RBAC, an immutable yield/quality-grade audit log and an on-prem or air-gapped option. Fleet learning shares model improvements, never your recipes.
Deposition, photolithography, encapsulation and anneal tools, Mura/AOI inspection and metrology, array and cell test, mass-transfer and repair stations, robotic handling and MES — vendor-neutral via SECS-GEM/HSMS where the tool supports it, with REST and OPC-UA bridges elsewhere.
Every pixel, perfectly uniform.
Ask us the one that is not here
We would rather answer it now than have it derail a pilot later.