Question 1

Does it work with scanned PDFs that aren't real text?

Accepted Answer

Yes — the system runs OCR on import, though accuracy depends on scan quality.

Question 2

Do I have to reprocess everything when I add new documents?

Accepted Answer

No. New documents are added to the existing set — the system only indexes the files you add.

Question 3

Is this live with a real customer yet?

Accepted Answer

Not yet — it's still in development, with no customer or pilot running it yet. If you'd like to be an early tester, get in touch.

Question 4

Our factory lighting isn't consistent — does this still work?

Accepted Answer

Our on-site assessment team checks the actual lighting conditions first, and will recommend supplemental lighting for the inspection point if needed.

Question 5

We change product models often — does the model need retraining every time?

Accepted Answer

In most cases we fine-tune the existing model to recognize the new product, rather than starting from scratch.

Question 6

Do you have a real customer running this on their line yet?

Accepted Answer

Not yet. We're currently running internal tests with real parts from an actual production line, but no customer has it installed yet. If you'd like to be our first, get in touch.

Question 7

Do documents need to look identical every time?

Accepted Answer

The system can learn multiple layouts, but each layout needs enough training examples.

Question 8

How would I know if the system extracted something wrong?

Accepted Answer

There's always a review step before anything is saved for real early on, plus a way to flag corrections that improve accuracy going forward.

Question 9

Do you need to shut the machine down to install this?

Accepted Answer

In most cases, no — these are external sensors that don't require modifying the machine itself.

Question 10

Can you pull data from a machine with no electrical signal at all?

Accepted Answer

If the machine has no electrical or mechanical signal to tap into, our on-site assessment team will tell you directly that it can't be done.

Question 11

How long does it need to collect data before alerts become accurate?

Accepted Answer

Generally it needs at least several weeks of normal-operation data before the system starts distinguishing anomalies accurately.

Question 12

Does it work with motors of every size?

Accepted Answer

Our on-site assessment team checks whether the motor or machine has a suitable mounting point for the sensor before the project starts.

Question 13

Can this system identify which employee broke the rule?

Accepted Answer

The system focuses on detecting whether PPE is worn per the rules for that zone — it isn't designed to identify individual employees.

Question 14

Do I need an RTX 6000 Ada specifically?

Accepted Answer

No — we're currently testing on an RTX 6000 Ada, but the actual hardware spec can be scaled to your data size and the model you choose.

Question 15

Why not just use ChatGPT or a cloud API instead?

Accepted Answer

Because a lot of organizational data can't leave the building — customer data, contracts, trade secrets — and per-token API costs add up fast once staff start using AI every day. This system runs entirely on your own hardware: no data goes anywhere, and there's no per-query token cost.

Question 16

Are you locked into one specific model forever?

Accepted Answer

No. We use whichever open-source model fits the job best at the time, and switch to a better one as new ones come out — we're not tied to any single model.

Question 17

What kinds of test results does this work with?

Accepted Answer

The system learns from the examples an expert labels for it. It works for results that are distinguishable enough from an image — color, shape, a meter needle position — and needs enough real examples before training starts.

Question 18

How many example images do I need to get started?

Accepted Answer

More examples generally means better accuracy. The right number depends on how complex each result type is — we can give you a specific estimate once we talk through your actual case.

Question 19

What does it start at?

Accepted Answer

A site assessment starts at ฿25,000, credited back in full if you go ahead. A pilot project starts at ฿120,000. A full installation is quoted against scope once we have seen the real site — there is no single price that fits every factory.

Question 20

Is there a monthly cost?

Accepted Answer

Yes — support starts at ฿8,000 a month, covering maintenance and system updates under contract. The final figure depends on the number of cameras, the complexity of the job and the site conditions. We state it clearly after the assessment, with nothing hidden.

Question 21

Can we ask you to delete our company's data?

Accepted Answer

Yes — that is a data-subject right under Thailand's PDPA. Contact us through the channel in the privacy policy to exercise it at any time.

Frequently asked questions

Getting started

Technical capability

Pricing and contracts

Security and PDPA

Support after installation

Frequently asked questions

Getting started

Can we use cameras we already have?+−

How long until we see results?+−

Our factory lighting isn't consistent — does this still work?+−

Do you need to shut the machine down to install this?+−

Technical capability

Can this connect to our existing ERP/MES?+−

If we change what we manufacture, does everything need to be redone?+−

What can't the AI do?+−

Does it work with scanned PDFs that aren't real text?+−

Is this live with a real customer yet?+−

We change product models often — does the model need retraining every time?+−

Do you have a real customer running this on their line yet?+−

Do documents need to look identical every time?+−

Can you pull data from a machine with no electrical signal at all?+−

Does it work with motors of every size?+−

Can this system identify which employee broke the rule?+−

Do I need an RTX 6000 Ada specifically?+−

Are you locked into one specific model forever?+−

What kinds of test results does this work with?+−

How many example images do I need to get started?+−

Pricing and contracts

What does it start at?+−

Is there a monthly cost?+−

Security and PDPA

Where does the training data end up?+−

Why not just use ChatGPT or a cloud API instead?+−

Can we ask you to delete our company's data?+−

Support after installation

Do we need our own IT team?+−

What happens if the model gets something wrong?+−

Do I have to reprocess everything when I add new documents?+−

How would I know if the system extracted something wrong?+−

How long does it need to collect data before alerts become accurate?+−