AI in Electronic Equipment, Instruments & Components

BOE built an in-house deep-learning defect classification system for display panel lines, trained on AOI images in its data center and run on Intel Xeon edge servers with OpenVINO. Deployed across its factories, it reached 96% inspection accuracy and a 70% labor substitution rate.

Foxconn built Nurabot, a collaborative nursing robot, on the NVIDIA Isaac for Healthcare framework and is field-testing it at Taichung Veterans General Hospital. The page reports a 30% cut in nurse workload and 2–3 hours saved daily on transport tasks, with simulation-driven validation reducing deployment time by 40%.

Foxconn's Shenzhen Lighthouse factory runs 2,000 BOXiedge edge AI units — built on Hailo-8 and SynQuacer silicon delivering 26 TOPS each — in a three-tier cloud-edge-device architecture that compresses inspection response time from 300 ms to under 15 ms. The system has raised daily throughput from 120,000 to 960,000 units, replaced more than 3,000 manual inspectors, and saved an estimated ¥320 million in annual quality-inspection costs while driving PCB inspection from 120 s to 22 s per board and lifting mobile-screen yield from 95% to 99.9%.

At TCL CSOT's 540,000-square-metre Suzhou display panel factory, GeChuang Dongzhi's Tianshu AI runs a closed loop of automatic defect classification and repair, identifying more than 4,000 types of micrometre-level defects at over 95% accuracy. GeChuang says it has replaced more than 75% of manual inspection and runs more than 100 models at once on edge devices.

LG Innotek replaced traditional rule-based visual inspection with an Intel-based AI vision system across its precision component manufacturing lines, achieving 99.99% defect detection in production. The deployment combines Intel Core, Arc, and Xeon processors with the OpenVINO inference engine, delivering 3-4× cost efficiency versus prior hardware and enabling expansion into semiconductor substrate production at Gumi Factory.

Hexagon worked with AWS's Prototyping and Cloud Engineering team to replace manual digitisation of standard operating procedures for its AcceleratorKMS software with a serverless pipeline: Claude 3.7 Sonnet on Amazon Bedrock extracts each page into structured JSON, Step Functions and Lambda orchestrate processing, and a review interface keeps humans in the loop. Accuracy exceeds 90%.

Longcheer Technology, a smart device ODM manufacturer, has integrated multiple AGIBOT G2 robots into its tablet production lines, where they work alongside human operators. AGIBOT reports throughput of up to 310 units per hour, a success rate above 99% and production line integration completed within 36 hours.

SkyKelpie, a Queensland drone mustering company, took livestock herding algorithms developed in a simulation environment and prototyped them in real paddock conditions across south-east and north-west Queensland. The work is a collaborative project with Meat & Livestock Australia and the Queensland Department of Primary Industries.