AI-Powered Visual Inspection for Manufacturing

Connecting factory devices, inspection models, and quality data in one platform to help manufacturing teams monitor production and investigate defects.

Client

Exacto

Industry

Manufacturing

Solution

Enterprise IoT edge intelligence platform

Overview

Development of an Enterprise IoT Edge Intelligence Platform for AI-Powered Visual Inspection

Manufacturing teams need to identify defects early and understand what is happening across production lines. Manual spot checks, disconnected inspection records, and time-consuming reports can make that difficult. We built Exacto, an enterprise IoT edge intelligence platform that connects cameras, edge devices, programmable logic controllers (PLCs), and machine learning models with centralized monitoring and reporting. The platform brings device setup, model deployment, inspection results, and audit records into one workflow. It supports multi-camera configurations and both cloud and on-premise storage to accommodate different factory environments.

Project at a Glance

The Challenge

Manufacturing environments offered no single standard to build against.

The Solution

We delivered a full-stack IoT intelligence platform built around nine integrated modules.

Services

Technologies

Apache Solr, OPC UA, Modbus, AWS S3, On-premise NAS

Challenge

The Challenging Part

## Making Automated Inspection Practical on the Factory Floor Automating visual inspection requires more than a trained model. Teams also need a reliable way to configure equipment, deploy the right model, monitor production, and trace results.

Configuration Requirements | Factories used different cameras, edge devices, communication protocols, and PLC systems. Bringing them into a consistent setup process required flexibility. Time-Consuming Device Setup | Device commissioning depended on manual scripts and repeated configuration steps. Teams needed a guided process that made setup easier to complete and maintain. Limited Production Visibility | Operators needed current inspection results to investigate failures during production. Managers also needed access to quality trends without waiting for weekly reports. Growing Inspection Data | Images, timestamps, model outputs, and pass/fail results accumulated across production cycles. Finding relevant records became increasingly difficult as the dataset grew. Model Deployment and Traceability | The correct inspection model needed to be associated with the appropriate device and part number. Teams also needed a record of model versions, configuration changes, and user actions. Different Data Storage Requirements | Some deployments required cloud storage, while others needed inspection files to remain on-premise. The platform had to accommodate both approaches.

Inside the Product

A closer look

Solution

How the Solution Worked

## One Connected Workflow for Manufacturing Quality Inspection

Guided Device Onboarding | A step-by-step setup flow helps teams register edge devices, test connectivity, configure cameras, and assign inspection models. Edge-Based Inspection | Results are then made available to the central platform for monitoring, search, and reporting. Production Monitoring | Operators can view inspection images, pass/fail outcomes, and First Pass Yield (FPY) the proportion of inspected parts that pass on their first inspection. Centralized ML Model Management | A model library brings uploads, version tracking, and deployment assignments into one place.

Core Technologies

We used the following technologies to deliver this project.

Next.js Python Flask Tailwind CSS Redux Persist Apache Solr WebSockets Formidable Multer Nodemailer Docker Compose Caddy Prisma Studio AmCharts Typescript Docker Node.js

Results

Real Impact for the Business

## From Disconnected Tasks to a Coordinated Inspection Workflow

100% | inspection coverage across all production cycles 1M+ | inspection records in seconds Up to 6 cameras | per device managed through one unified form Device commissioning reduced | from multi-day manual scripting to a single guided session ML deployment errors eliminated | centrally version-tracked and audited Quality reporting shifted to live dashboards | real-time FPY data filterable by shift, part number, and model

Client Testimonial