Services
Computer vision and AI for companies: documents, automation and visual inspection
We build AI systems for repetitive work whose results can be measured, such as checking purchase orders that arrive by email or finding substandard material on a production line.
- Email received
- Documents read
- Master data check
- Draft quotation
- 5Staff review
Work we handle with AI
-
Reading documents and email
The system reads incoming documents, such as purchase orders and technical drawings, extracts the data needed and matches it against specifications or the data in your systems.
Checked data and a list of documents that staff need to review.
-
Automating repetitive work
Steps now done by hand, such as copying data from email into the ERP or preparing a quotation, are run by the system under rules you set.
The status of each document and a log of the steps the system ran.
-
Visual inspection on production lines
A camera captures images of material or products on the line, and the system marks anything that does not meet the standard (NG, no good) before it moves to the next process.
A marker on the parts that staff need to re-check.
-
Video and CCTV analytics
For surveillance cameras in bank branches, ATMs and stores, we use MocengVision, a platform built by Teguh Hardiansah and Nabil Basyuni. Details are on the AI CCTV page.
Visitor and queue counts, event notifications and routine reports.
Where computer vision is used
Computer vision is the branch of AI that recognises objects, activities and events in images or video. Factories use it for visual inspection. In bank branches and stores, the same technology counts people and measures queues from CCTV video.
MVP Technologies is a technology solutions and integration company in South Jakarta. Besides the AI model, we can supply the cameras and servers and connect the system to the applications you already use.
- Camera field of view
- Laser dot on NG material
Example visual inspection flow: a camera on the production line, detection of NG material, then a marker on the part to be checked.
AI projects at PT TT Metals Indonesia
Core team experience from 2023, before PT Marsada Visi Pranata was established. Both projects were delivered by PT Alkeira Kreasi Indonesia, where Teguh Hardiansah, now MVP's Director, was CTO from 2021 to 2023.
-
2023
PT TT Metals Indonesia
AI agent for purchase order verification
PT TT Metals Indonesia, part of the Toyota Tsusho group, processes steel for the automotive industry in Karawang. The system read purchase orders and technical drawings received by email, then checked the part specifications. The model ran on an NVIDIA GPU server at the plant. Processing time was measured in the company's Microsoft Dynamics ERP, from the email arriving to the quotation being sent.
About 1,000POs per month1–3 daysprocessing time before< 15 minutesprocessing time after- Delivered by
- PT Alkeira Kreasi Indonesia
- Status
- Completed
-
2023
PT TT Metals Indonesia
Raw material inspection with an AI camera
The inspection system detected NG (substandard) material on the production line before it was processed. A high-resolution camera was modified for inspection, and a red laser dot marked the part to be checked. One camera was installed on the line.
1camera on the production line20 hoursof operation per day- Delivered by
- PT Alkeira Kreasi Indonesia
- Status
- Completed
AI and computer vision project steps
Defining the task and how to measure it
We write down one clear task and how its result will be measured, for example the processing time from document receipt to completed work.
Checking the data
We collect sample documents, photos or video from daily work, including difficult cases such as blurred scans or small defects.
Preparing the model and equipment
We choose the model, prepare the server or camera, and connect the system to existing applications such as email and the ERP.
Testing in the working environment
The system runs on real documents or on the real production line. Its results are compared with staff checks, and the rules or model are corrected before full use.
Handover and monitoring
We train users, hand over documentation and monitor the results once the system is live.
What you need to prepare
-
Sample data
Documents, photos or video from daily work, including problem cases.
-
Inspection rules
The rules your staff apply today, including tolerances or the list of specifications that must match.
-
System access
Access to email, the ERP or other applications the system needs to read or update, with approval from your IT team.
-
Where the model runs
A decision on whether the model runs on a server at your site or in the cloud, in line with your company's data policy.
-
Reviewing staff
Staff who review the system's results during testing and help decide when it goes into full use.
FAQ
Questions about AI and computer vision
Does our company data have to go to the cloud?
No. The model can run on a server at your site. In the PT TT Metals Indonesia project, it ran on an NVIDIA GPU server at the plant.
How much sample data is needed?
It depends on the task and on how varied the documents or materials are. We decide after seeing your sample data.
Are AI results used without human review?
Not necessarily. The system's output can be set up as a suggestion that takes effect only after staff approve it, with every change logged. We use this approach in the application we are developing for an outsourcing company.
Can we use the cameras already in our factory?
Yes, if their resolution and position are good enough to show the defects you are looking for. In the TT Metals project, a high-resolution camera was specially modified for inspection.
Can the system enter data directly into our ERP?
Yes, if your ERP provides an API or another integration method approved by your IT team. We check this during preparation.
How accurate is the AI system?
Accuracy depends on the quality and variety of the sample data, the camera position and lighting for inspection, and how clear the inspection rules are. We measure it with your staff during testing in the working environment, before the system goes into full use.
Related pages
Discuss the work you want to automate
Describe the work you want to automate and prepare a few sample documents or photos. We will assess whether AI suits the task and how to measure the result.