Line 2 · Pouch sealer
184,220packs inspected this shift
- Rejects
- 212
- Reject rate
- 0.12%
- Decision time
- 31 ms
Edge AI · Computer vision · Industrial IoT
Your plant has been trying to tell you something
Perflens puts cameras, sensors and AI at the edge of your operation so you see the defect, hear the failing bearing and catch the wasted kilowatt — before they become scrap, downtime or a bill.
Three things worth watching
What your line makes, how your machines run, how your building behaves
Most plants buy three systems from three vendors for these jobs, and none of them talk to each other. We build one.
What your line makes
AI visual, thermal and multispectral inspection that sees the scratch, open seal or short shot at full line speed and pulls the part before it ships.
Read more
How your machines run
Vibration, ultrasound, current and thermal monitoring that turns a failing bearing into a scheduled work order weeks before it becomes a breakdown.
Read more
How your building behaves
Energy, compressed air, HVAC and occupancy monitoring that shows where a facility spends money when nobody is looking.
Read moreHow it works
Sense, decide, act, keep it sharp
The Perflens Edge Stack is four layers you own outright. Nothing in it needs the internet to do its job.
-
Sense
Smart optical sensors and condition nodes capture images, heat, vibration, ultrasound and electrical signatures where the problem actually happens.
-
Decide
A fanless edge gateway runs the models beside the equipment and makes a decision in milliseconds — reject, alert, or carry on.
-
Act
Decisions become reject pulses, stack lights, operator messages and CMMS work orders, with an image or a trend attached to every one.
-
Keep it sharp
Models drift as materials, seasons and machines change. Care monitors drift, verifies alerts and retrains before accuracy slips.
Why the edge
Decisions belong next to the machine, not in someone else's data center
A reject gate has a few milliseconds to act. A bearing alarm is useless if it waits for a WAN link that went down on Saturday night. And many plants simply won't send production images to a public cloud — nor should they have to.
So every Perflens system runs on hardware at your site. Inspection decisions take under 40 ms from frame to output. Monitoring keeps working when the network doesn't. Your images, trends and trained models stay on your premises, and the models are yours to export.
The cloud is optional: useful for comparing sites, never required for a single decision.
Services
Thirteen services, one accountable integrator
From a single inspection cell to a fleet of monitored sites — designed, installed and looked after by the same engineers.
Vision & Inspection
AI visual inspection, retrofits for existing cameras, thermal and multispectral monitoring, synthetic data and an on-device vision-language copilot.
Read moreEdge Hardware
Compute selection and integration, custom optical sensors, embedded firmware and electronics for plants and equipment makers.
Read moreIoT Monitoring
Predictive maintenance, equipment performance and OEE tracking, and facility and energy operations monitoring.
Read moreLifecycle
Perflens Care keeps models and alerts accurate; Edge & OT Security keeps every device hardened and patched.
Read moreCommitments, not claims
Numbers we put in writing
3 weeks
Feasibility Sprint from samples to a written go/no-go
< 40 ms
Typical frame-to-output decision time on a Gatepoint
0 bytes
Data that must leave your site for the system to work
2 days
On-site anywhere in the contiguous US with Care Verified
Specifications and service commitments from our published pricing and SLAs. Your Feasibility Sprint produces numbers for your own parts and assets.
Published pricing
Start small, with a fixed price
Every engagement can begin with a fixed-price step. The full price list, including subscriptions and capital purchase, is on the pricing page.
Feasibility Sprint
$9,500
fixed · 3 weeks
Your samples in our optics lab, a baseline model and a written go/no-go with modeled ROI.
- Lighting and camera study on your real parts
- Baseline model and measured accuracy
- 100% credited if you proceed within 90 days
Proof of Value
$24,000
fixed · 6 weeks
A loaner system on your line, measured against success criteria agreed in writing.
- One inspection station or up to 25 assets
- Loaner hardware included
- 50% credited to a subscription
Inspection-as-a-Service
$2,950
per station per month
Cameras, compute, software, Care Essentials and hardware replacement in one monthly fee.
- No capital request needed
- Hardware refresh included
- Shared-savings option after a measured baseline
Example engagements
What a deployment looks like
Representative scenarios built from typical deployment parameters, with modeled results — so you can see the approach before we model yours.
Heat-seal inspection at 180 packs per minute
Low-angle red light and a Gatepoint G2 catch leaking snack-pouch seals that rule-based vision kept passing.
Read more
Continuous thermal on switchgear and battery storage
Fixed radiometric sensors flag loose connections weeks before the next periodic IR survey would have.
Read more
Condition monitoring on 38 corrugator drives
Wireless vibration nodes and analyst-verified alerts turn breakdowns into planned work.
Read moreHow we work
Every step has an exit
You can stop after any step with something useful in hand. We would rather tell you "no" in week three than in month nine.
-
Discovery call
We ask about the defect, the asset or the building — and what being wrong costs you.
-
Feasibility Sprint
Lab work on your samples and a written go/no-go with modeled ROI.
-
Proof of Value
A loaner system on your line, measured against criteria agreed up front.
-
Deployment
Installation, integration, acceptance testing, training and documentation.
-
Care
Drift monitoring, verified alerts and retraining so results hold up.
Field Notes
Lessons from the plant floor
Practical writing from our engineers on vision, edge hardware, reliability and the economics of all three.
Vision-language models on the production line: where they help and where they hallucinate
A VLM can tell an operator why a pouch was rejected and find every similar reject from last month. It should never be the thing that decides to reject it.
Mara Lindqvist · October 28, 2026
What a minute of downtime really costs: a calculation you can defend
Most downtime cost figures fall apart at the first question from finance. A conservative method built on contribution margin, idled labor and restart scrap, worked through with hypothetical numbers.
Grace Nakamura-Hill · October 20, 2026
An IEC 62443 hardening checklist for edge devices
IEC 62443 is a long series with a short practical core. This is the checklist we work from when hardening an edge device, mapped to the standard, plus what the EU Cyber Resilience Act now expects of device makers.
Alicia Brandt · October 6, 2026FAQ
Questions we hear often
What does Perflens actually do?
We design, install and look after systems that watch production lines, machines and buildings — cameras and sensors, edge computers that run AI models beside the equipment, and software that turns what they see into rejects, alerts and work orders. One integrator, accountable for the whole chain.
Do we need to send data to the cloud?
No. Every Perflens system makes its decisions and stores its data on hardware at your site. Remote support uses an outbound connection you control and can switch off. Multi-site cloud dashboards are optional.
Can you work with the cameras and sensors we already have?
Often, yes. Vision Retrofit adds AI to many existing industrial cameras, and Gatepoint speaks OPC UA, MQTT, Modbus, EtherNet/IP and PROFINET, so it can read the machines and PLCs you already run.
How do we start without a big commitment?
With a 45-minute discovery call, then a fixed-price $9,500 Feasibility Sprint that ends in a written go/no-go. If the answer is no, you have spent three weeks and a known amount — not a year.
Show us the problem you keep working around
A 45-minute discovery call with an engineer — not a sales script. We'll tell you whether it's a vision problem, a sensing problem, or neither.