Pre-launch · building with design partners

We build theoperations layerfor appliances.

Duromex is an independent physical-AI product company. We build one thing: an autonomous operations layer that sees, decides and acts across appliance manufacturing lines — from steel blank to packed pallet.

1
industry, deliberately
7
stage agents
5
appliance categories
$22B
addressable market

Every unit, first-pass perfect

Appliance manufacturing runs on thin margins and unforgiving tolerances. A foam void becomes an energy-rating failure. A seated connector that is almost seated becomes a warranty claim in eighteen months. A pinhole in enamel becomes a downgrade.

These losses are not caused by a lack of data. Plants are full of data. They are caused by nothing acting on it fast enough, at every station, on every unit, every shift.

Duromex exists to close that loop — to make the line perceive, reason and act within takt time, with a full evidence trail behind every decision.

What we are not

  • Not a consultancy — we ship a product, not billable hours
  • Not a reseller — the perception, reasoning and control stack is ours
  • Not a dashboard — we act on the line, not just report on it
  • Not horizontal — the models are appliance-specific by design
  • Not a systems integrator — we deploy the same runtime everywhere

Why appliances, and why now

01

The industry is squeezed

White goods is a high-volume, low-margin, regulation-heavy industry where a point of first-pass yield is worth more than most software line items.

  • Thin margins, high unit volumes
  • Energy-rating and safety regulation
  • Warranty exposure measured in years
02

The data already exists

Presses, enamel lines, foamers, torque tools and leak testers already emit rich telemetry. Almost none of it changes what happens next.

  • 20–60 sensor streams per line
  • Serialised units with real genealogy
  • Test outcomes that could close the loop
03

The models finally work

Physical AI — vision, sensor fusion, simulation and reasoning at the edge — crossed the threshold where a station decision can be made in under 100 milliseconds.

  • Edge GPU inference within takt
  • Simulation good enough to gate change
  • Reasoning models that can cite evidence
04

Nobody built it for this industry

Generic MES vendors know factories but not foaming. Vision vendors know cameras but not enamel cure. Neither closes the loop on the appliance itself.

  • Vertical models beat horizontal ones
  • The unit, not the station, is the object
  • Attribution across stages is the hard part

What we believe

These are engineering positions, not slogans — they show up in the product.

Trust is earned in shadow mode

No system should be allowed to write to a press before it has proven it beats the human baseline while writing nothing.

Every action needs an evidence chain

If an agent cannot show why, an engineer cannot defend it and an auditor cannot accept it.

Vertical depth beats horizontal breadth

An agent that understands polyurethane cure behaviour is worth more than a general anomaly detector.

The edge is the product

If the decision depends on a cloud round-trip, it is not a station decision — it is a report.

Customer IP stays customer IP

Recipes, defect libraries and quality data never leave the tenant and never train a competitor's model.

Reversibility over cleverness

Any change an agent makes must be observable, bounded and instantly revocable.

Blank to pallet, one continuous system

We chose to cover the whole build rather than one station, because the defects that cost the most are created in one place and discovered in another.

Form & stampWeld & seamEnamel & paintFoam & insulateWire & assembleLeak & function testRate & pack
PASS · A+++ 1 2 3 4 5 6 7

Where we are and where we are going

Stated as plans, not promises.

Now

Design-partner cohort one

Instrumenting real lines across refrigeration, laundry, cooking and air conditioning with written success metrics.

  • Shadow-mode baselines under way
  • Connector families hardening
  • Agent evaluation harness in CI
Next

Advisory mode at pilot scale

Recommendations under human approval, with attributable deltas against published baselines.

  • Twin gating on change
  • Exception escalation workflows
  • First published case studies
Then

Scoped control on low-risk workflows

Bounded, revocable write authority where the evidence supports it.

  • Per-line authority envelopes
  • Canary and rollback in the field
  • Multi-line rollout inside a plant
Later

Multi-plant fleet learning

Cross-site improvement without cross-tenant data leakage.

  • Fleet benchmarking within a tenant
  • Recipe portability between plants
  • Enterprise-wide assurance reporting

The opportunity we are building into

$22B

total addressable market

$5.5B

serviceable available market

$300M

serviceable obtainable market

12%

market growth rate

Market sizing is our own modelling of appliance-manufacturing operations software and physical-AI spend. Treat it as an estimate.

The facts, plainly

What an investor, partner or programme reviewer usually asks first.

Company type
Independent product company — we build and license our own software, not client projects
Category
Physical AI for discrete manufacturing, specialised in major appliances and white goods
Headquarters
London, United Kingdom
Stage
Pre-launch, building with design partners on real production lines
Funding
$1.5M seed round, backed by 99SA Ventures
Leadership
Kabeer Iqbal, Co-Founder & CEO · Maria Li, CTO
Product
Edge runtime plus cloud control plane, deployed as one repeatable pattern per plant
Revenue model
Per-line and per-factory subscription with enterprise agreements — see pricing
IP position
Proprietary appliance perception models, control envelopes, twin and assurance layer
Deployment
AWS control plane plus on-premise factory edge; the control path has no cloud dependency

The people running Duromex

A small team in London, working on real appliance lines with design partners.

Kabeer Iqbal
Co-Founder & CEO

Kabeer leads Duromex’s commercial and product direction: which line goes first, how the success metric is written down, and who at the OEM signs off on it. He works with appliance manufacturers to move forming, foaming, assembly and test onto closed-loop AI, one measured line at a time.

Maria Li
Chief Technology Officer

Maria leads engineering: the factory-edge runtime, the vision and sensor-fusion models that decide within takt, the line-and-product digital twin, and the bounded, revocable write-back into line controllers. She is why every agent earns trust in shadow mode first and carries an evidence trail behind every action.

What partners tell us

“Our end-of-line functional testers told us what failed. They never told us which press stroke or foam shot caused it. That link is the whole product.”

Plant Director
Refrigeration OEM · design partner Illustrative

“Foam voids are invisible until the energy rating comes back wrong. Predicting density distribution before the cabinet cures is the part I could not buy anywhere else.”

Process Engineer
Cabinet line · design partner Illustrative

“If it writes to the press, it needs an audit trail a quality auditor accepts. Duromex started there instead of bolting it on.”

Quality & Test Engineer
Laundry OEM · design partner Illustrative

Segments we build for

Refrigeration OEM · EU
Laundry OEM · North America
Cooking OEM · Turkey
Dishwashing OEM · Italy
Air-conditioning OEM · Asia
Tier-1 cabinet supplier

Design-partner programme, 2026. Duromex is pre-launch — partner identities are withheld under NDA and shown by segment, not by logo.

Common questions

Are you a consultancy or a systems integrator?

Neither. Duromex licenses a product. The same edge runtime, agents and control plane deploy at every customer; integration work is configuration, not bespoke development.

Why not target all discrete manufacturing?

Because the value is in the vertical depth. Understanding foam cure, enamel film build and refrigerant leak signatures is precisely what a horizontal platform cannot do.

What happens to our data?

It stays in your tenant. Recipes, defect libraries and quality records are never used to train models for anyone else. See security.

How do you make money?

Per-line and per-factory subscriptions, plus enterprise agreements. Pricing is published on the pricing page.

Build the loop with us.

We are looking for appliance manufacturers who want a written success metric and an honest baseline — not a demo.

Duromex is pre-launch. Figures shown are design-partner targets and modelled economics, not audited results. Ask us for the methodology.