01Logistics & Supply Chain
Plan the route, the takt and the truck in one place.
A deterministic planning engine for everything that moves — inside the plant and beyond the gate. It holds your capacity limits, delivery windows, priority rules and consolidation logic, and recomputes the entire plan in under two seconds when an input changes. Same inputs, same plan, every time.
Milkrun & takt planning
Internal routes and line-side replenishment cycles, balanced against takt. The constraint model is built from your plant's rules, not fitted to a template.
Transport management
Inbound and outbound shipments, carrier assignment, load consolidation, dock and delivery-window scheduling, freight cost tracked against plan.
Freight rate intelligence
Continuous analysis of maritime, energy and geopolitical signals, scoring pressure on each corridor before spot rates move — so you re-route, pre-book or hedge on notice rather than on news.
Auditable by design
Raw data cleaned and reconciled on the way in. Every planning run kept, so any decision can be reconstructed months later.
Ten weeks from kick-off to go-live. Passed 8 of 8 independent acceptance tests on the plant's own operating data.
02Quality & AI Inspection
Every part inspected. Every decision on the record.
Camera inspection on the line replaces sampling at the end, and the finding goes straight into the quality record against the part it belongs to. Detection is half the product. What happens to the defect afterwards is the other half.
Inspection on the line
Every unit checked at production speed. Surface, weld, assembly and dimensional defects classified and logged with part-level attribution.
Trained on your line, not someone else's
The model learns your defect library, your tolerances, your lighting — and is retrained as the process changes. A model trained on a general dataset finds general defects; yours finds the ones that actually come back from your customers.
AI you can put in front of an auditor
Every decision recorded with its confidence score. Detection rate and false-reject rate published per model, per line. Borderline cases route to an operator, and that sign-off becomes part of the record.
Quality management, not just detection
SPC charts on live data, non-conformance reports raised from the defect itself, corrective actions tracked to closure, PPM trends and defect Pareto by line, shift and station.
03Traceability
Answer a recall question in minutes, not a week.
Complete genealogy for every unit you ship: which raw batch, which machine, which operator, which shift, which inspection result, which pallet, which customer. Built from data the other systems already produce.
Forward and backward in one query
Start from a defective unit and find every other unit that shares its cause. Start from a suspect raw batch and find everywhere it went. Scope a recall to the units actually affected instead of the whole month's output.
Genealogy across the whole build
Component, sub-assembly and finished unit linked through production, including rework and re-inspection history.
Evidence your customer will accept
Export packs formatted for automotive, electronics and export-market requirements, with the audit trail attached.
This is the system that gains most from a shared foundation. It reads material flow, quality and equipment data as one record — not three exports that someone has to reconcile by hand at the worst possible moment.
04OEE & Andon
The real reason the line missed its number.
Availability, performance and quality computed from machine data — not typed into a spreadsheet at the end of the shift. The number stops being flattering, which is the entire point of measuring it.
Losses named, not totalled
Every lost minute attributed to a cause and a machine: changeover, minor stops, speed loss, startup rejects. A single OEE percentage tells you there is a problem. The loss tree tells you where to stand tomorrow morning.
Andon that measures the response
Station-level call for help with escalation timers. Response and resolution times are recorded, so the improvement conversation moves from "the line stopped" to "it took eleven minutes for anyone to arrive."
Comparable across lines, shifts and sites
The same definitions everywhere, so a shift comparison means something and a plant comparison is not an argument about methodology.
No new sensors
Computed from connections the foundation already carries.
05Equipment Telemetry & Predictive Maintenance
The machine tells you before it stops.
Device streams filtered for noise, scored against tuned thresholds, ranked by risk, and turned into work orders scheduled inside shift-change windows.
- —Telemetry pipeline proven at 100,000+ connected devices
- —Detection rate and false-alarm rate published per model — measured, not asserted
- —Risk-ranked failure list, not an undifferentiated alert stream
- —Maintenance scheduled into planned windows instead of emergency call-outs
06Data Security
The drawing does not leave the building without a signature.
Your process knowledge lives in files: drawings, BOMs, work instructions, supplier terms, customer data. Most of it leaves the plant legitimately every day, so the problem has never been stopping that — it is knowing which file went where, on whose authority, and being able to show it afterwards.
DocGuard — approval before the file moves
Controlled documents are classified at rest and watched at the endpoint. When one is attached to an email, sent through Zalo or uploaded to a website, the send is held and routed to the named approver, who sees the file, the destination and the stated reason before anything moves. Approved, it goes with that decision attached to the record. Declined, it never leaves the machine.
Built for plants that do not run on Microsoft
We surveyed the commercial DLP market before building this. Every manager-approval workflow we could find is tied to Microsoft Exchange or Microsoft 365 — which excludes any plant that does not run its mail on Microsoft. DocGuard sits on the endpoint and on the send path, so it works with the mail client and the messaging app your staff already use.
Encryption, with the keys on your side of the table
AES-256 at rest, TLS 1.3 in transit, and field-level encryption for personal and commercially sensitive columns. Keys live in your KMS or HSM, under your rotation policy and your separation of duties. We do not hold a key that opens your data — a property you can verify rather than a promise you have to accept.
Database security
Access granted by role, down to the column. Production data masked in test and development copies. Privileged-account activity monitored, with bulk reads and out-of-hours exports flagged. Every read and every change to a sensitive table written to an audit log the database administrator cannot quietly edit.
Written for the rules that apply from 2026
Vietnam's personal data protection framework — Luật Bảo vệ dữ liệu cá nhân and Nghị định 356/2025/NĐ-CP — took effect on 1 January 2026. Data inventory, lawful basis, retention limits, deletion on request, cross-border transfer records and incident reporting are obligations that sit with the enterprise, not with its vendors. The system is what turns them from a policy document into something you can demonstrate on the day you are asked.
Data security is graded on the day something goes wrong. What counts then is not the policy that was written, but whether the log can say what left, when, and who approved it.
07EcoCheck
Your emissions report, computed from the plant's own data.
Scope 1, 2 and 3 under the GHG Protocol and ISO 14064-1, calculated continuously from meter and process data rather than reconstructed from spreadsheets at year end. Every figure traces back to the reading that produced it.
- —Emission factors maintained against IPCC, IEA, DEFRA and national grid sources
- —Real-time dashboard across sites, with permissions by plant
- —Report exports configured to the format your regulator, parent group or CBAM filing requires
- —Full audit trail from meter reading to published figure
In production with enterprise clients across Vietnam and Southeast Asia.