Trust before autonomy
Grounding, audit, human-in-the-loop and graduated control come first, always.
Company
Keep the craft on the floor
Foldeon started on a plant floor, not a whiteboard, watching lines reject stack after stack because a corrugator warped a lighter board, a press drifted off brand color, or a folder-gluer skipped a glue line no one caught until the boxes failed.
The machines could run beautifully, but only when a shrinking group of veteran operators were there to feel the drift and correct it.
When those people retire, the quality walks out with them. Foldeon was founded to keep that craft on the floor by encoding it in software that senses, predicts and acts.
Why a layer, not a box
One layer across machines that barely talk
No single-machine control can win converting, because the machines come from different vendors and interact, fix warp and you can move color; speed up folding and you can skip glue. The opportunity was a layer that sits across all of them and learns from every box.
- Closes the loop from perception to control
- Sits across corrugator, press, die-cutter and folder-gluer
- Learns from every run and every correction
- Vendor-neutral by design
The stance
What we believe
The principles behind every design decision.
A compounding data moat
Every supervised correction trains the system. The moat is the dataset, not any one model.
Land narrow, expand relentlessly
Win one wedge with undeniable ROI, then earn the next line, module and site.
Outcomes over dashboards
We are measured on waste, makeready, claims and OEE, a system of action, not reporting.
Roadmap
From wedge to autonomous operations layer
0–6 months
Design partners & MVP
Onboard 3–5 design partners on the wedge workflow. Ship connectors, the agent loop, review console and quality-grade audit log. Prove ROI in shadow then assist mode; establish security and quality baseline (SOC 2 Type I in progress).
6–12 months
Paid conversion & second module
Convert design partners to paid and reach early revenue. Graduated autonomy for low-risk lines and hit-quality-before-run twin. Add a second plant module; SOC 2 Type II, SSO/RBAC, first enterprise logos.
1–3 years
Own the full loop
Own corrugate → print → die-cut → fold-glue → palletize end-to-end across vendor-siloed machines. Multi-segment GTM, OEM and inspection-vendor channels, international expansion, benchmark data products. Scale to $36M–$75M ARR.
3–10 years
The operations layer for converting
Become the autonomous system of action for the function globally, a platform and API ecosystem across corrugated, folding carton, labels, flexible packaging and rigid boxes. Every box and carton made on Foldeon.
3–5Design partners, year one
$36–75MARR target (1–3 yr)
$24BTAM
Multi-billionLong-term vision
The moat
A cross-machine dataset no one else can assemble
Because Foldeon runs across every stage and vendor, it builds a converting-quality-and-waste dataset that single-machine incumbents structurally cannot, and every supervised correction deepens it.
- Cross-machine, cross-vendor perception and outcomes
- Per-plant and per-craftsman versioned memory
- Synthetic augmentation for rare defects
- Compounds with every line and plant added
The quality was walking out the door with the people who could feel the drift. We built Foldeon to keep it on the floor.
Building on the NVIDIA physical-AI platform
JetsonIsaacMetropolisHoloscanOmniverseTriton · NIMcuOpt
Answers
Questions converters ask
What stage is Foldeon at?
Foldeon is an early-stage, independent physical-AI startup building original plant-edge autonomy software for packaging converting. Roadmap milestones on this page are aspirational design targets.
Are you hiring?
Yes. We hire ML, robotics, controls and field engineers who want to work on physical AI where the loop closes on real machines. See careers.
Where are you based?
Foldeon works close to plant floors. Reach out via the contact page to talk to the team.
Build the operations layer for converting
Whether you run a plant or want to build the platform, we’d like to talk.