AI that runs inside your factory
Not on someone else's cloud
We design, deploy and operate edge, local and open-source AI from site assessment through live operations, choosing the processing boundary around your data, timing, connectivity and long-term support needs.
- Customer data processed on-site
- Timeline set after the site assessment
- Existing cameras and PLCs assessed first
We start from the problem you actually have, not from the technology
Search company documents
Manuals, contracts, and SOPs run into the tens of thousands of pages — finding anything means opening files one by one
Local LLM · OCR · RAG
Read the detailInspect quality with a camera
Defects reach the customer because inspectors get tired on the night shift
Edge Vision · Custom Model
Read the detailAutomate office work
Staff spend 3 hours a day keying in data from purchase orders
Local LLM · Workflow Automation
Read the detailConnect machines and sensors
A 20-year-old machine has no data port — you don't know how it's running
Raspberry Pi · Modbus · MQTT
Read the detailPredictive maintenance
A motor's failure is only discovered once the line has already stopped
Vibration · Acoustic ML
Read the detailWorkplace safety
Someone has to walk around checking PPE compliance every shift
Edge Vision · Zone Detection
Read the detailSee what the system sees
Simulated example, not live output from a real system.
How far does your data actually travel?
Hardware we install
Processing hardware price only, excludes enclosure and installation · IP65/IP67 dust/water rating, operating temperature range, and PoE power depend on the model chosen.
| Device | Performance | Best for | Hardware price range |
|---|---|---|---|
| Raspberry Pi 5 + AI HAT | 13 TOPS | IoT sensors, counting, light workloads | ฿8,000–15,000 |
| Jetson Orin Nano Super | 67 TOPS | Vision inspection, 1–4 cameras | ฿25,000–40,000 |
| Jetson AGX Orin | 275 TOPS | Multiple cameras at once, large models | ฿90,000–150,000 |
| Workstation RTX 4090/5090 | — | On-premise document LLM systems | ฿150,000–300,000 |
- Device
- Raspberry Pi 5 + AI HAT
- Performance
- 13 TOPS
- Best for
- IoT sensors, counting, light workloads
- Hardware price range
- ฿8,000–15,000
- Device
- Jetson Orin Nano Super
- Performance
- 67 TOPS
- Best for
- Vision inspection, 1–4 cameras
- Hardware price range
- ฿25,000–40,000
- Device
- Jetson AGX Orin
- Performance
- 275 TOPS
- Best for
- Multiple cameras at once, large models
- Hardware price range
- ฿90,000–150,000
- Device
- Workstation RTX 4090/5090
- Performance
- —
- Best for
- On-premise document LLM systems
- Hardware price range
- ฿150,000–300,000
How it works
On-site assessment
1 week
We visit the actual site, measure lighting conditions and belt speed, review existing systems, and send a report on what's feasible and whether it's worth it.
This step has a fee, fully credited if you continue with us.
Real-world pilot
2–4 weeks
We install a test system at your actual site and agree on clear success criteria before starting.
Full installation
4–8 weeks
We install the complete system, connect it to your existing systems, train your team, and hand over documentation.
Ongoing support
Monthly
We check system health, retune the model when your production line changes, and provide a clear response-time SLA.
Choose the path that fits
From site data to a decision
Camera, sensor or document → Edge or local inference → Human approval → ERP/MES/workflow → Evidence and improvement
See industry contextHow we work
Edge-first when timing matters
Put decisions close to the source so a site is not waiting on a network round trip when response time matters.
Local-first when privacy matters
Keep business data inside the agreed boundary, with update and remote-support exceptions documented separately.
Open by design
Choose the model, runtime and hardware for technical and business reasons, with limits and replacement paths recorded.
Evidence we are still testing
Last reviewed 12 September 2026
Internal tests are not a customer guarantee. Figures appear only when the method and conditions are publishable.
Internal test
Detecting sink marks and flash on injection-moulded plastic parts
Dark, glossy parts — the condition that makes vision systems fail most often — so we designed the LED bar lighting and the camera angle before training any model.
See method and limitationsInternal test
Asking 4,200 pages of manuals questions the technicians wrote themselves
The 120 test questions came from the maintenance technicians, not from questions we picked because we could answer them. The documents include both digital files and skewed scans.
See method and limitationsInternal test
Tracking livestock growth with a home-built stereo camera, instead of one fixed scale
Two CCTV cameras plus a Raspberry Pi 5, calibrated into a stereo pair as a cheaper alternative to a LiDAR camera. It photographs animals on a scale to convert into volume and weight — but there's an unsolved statistics problem: a single fixed scale only captures 100-200 animals a day out of a pen holding 5,000-10,000, which is too small a sample by statistical standards.
See method and limitationsRead on in the Knowledge Centre
Enter the Knowledge Centretechnical-knowledge
What Edge AI is, and how it differs from IoT and Cloud AI
A decision guide for placing Edge AI, IoT, Cloud or hybrid components, including the operating responsibilities that remain after installation.
Read articleexecutive-explainer
Local AI or Cloud AI: an organisation decision table
Compare local, cloud and hybrid using data boundaries, latency, cost, connectivity and operating ownership instead of assuming one option is always best.
Read articletechnical-knowledge
Private RAG: a document-search architecture that keeps data in the boundary
Trace the flow from OCR to an answer, including access control, citations, versioning, audit and the limits to test before production.
Read articleBuild it, then keep it dependable
Pricing and engagement models
Price depends on camera count, task complexity, and site conditions. We give you a clear quote after the assessment — no hidden costs.
| Package | Starting price | Duration | What's included |
|---|---|---|---|
| On-site assessment | From ฿25,000 | 1 week | Feasibility report + ROI estimate + required specs |
| Pilot project | From ฿120,000 | 2–4 weeks | A working system on-site with agreed success criteria |
| Full installation | Quoted based on scope | 1–3 months | Complete system + training + documentation + warranty |
- Package
- On-site assessment
- Starting price
- From ฿25,000
- Duration
- 1 week
- What's included
- Feasibility report + ROI estimate + required specs
- Package
- Pilot project
- Starting price
- From ฿120,000
- Duration
- 2–4 weeks
- What's included
- A working system on-site with agreed success criteria
- Package
- Full installation
- Starting price
- Quoted based on scope
- Duration
- 1–3 months
- What's included
- Complete system + training + documentation + warranty
Monthly maintenanceFrom ฿8,000/month
Start with three short questions
Your answers only suggest a path. They do not calculate ROI or guarantee an outcome.
- Where in the workflow does the problem happen?
- Is the data image, document, audio or sensor data?
- Do you want to explore, design, pilot or operate an existing system?
Frequently asked questions
Tell us about the problem you're facing
A free 30-minute conversation. If we think your problem isn't a good fit for AI yet, we'll tell you directly.
30 minutes
Explain the problem, existing systems and on-site constraints.
A clear boundary
Know whether to pilot, design, collect more evidence or pause AI.
No vague promises
Advice tied to evidence, a test method and the person who will operate it.