Edge AI Deployment on NVIDIA Jetson
Workforce Development Agency, Ministry of Labour (iTVET) · Apr 2026
Running YOLO in the cloud as offline analysis is one thing; putting it on the line for real-time inference is another. This is the second skill.
🌐 改用繁體中文閱讀NVIDIA JetsonEdge AIYOLO InferenceEmbedded DeploymentIIoTReal-time Inference
📝 Summary
“AI in the factory” sounds like the future, but the real barrier is not model accuracy — it is where inference runs. This certificate is about that barrier.
🎯 Problem
Analysing YOLO output in the cloud is a different problem from making judgements on the line in real time. Factory floors have bandwidth limits, latency requirements, connectivity risk and thermal constraints that make cloud inference impractical in many situations.
🛠️ Approach
- Completed hands-on edge AI development and deployment in the iTVET 'NVIDIA Jetson' course (certificate awarded 2026-04-14)
- Evaluated the feasibility and resource constraints of deploying models on edge hardware for in-line inference
- Connected this to the SiC wafer defect-detection thesis, moving defect judgement from offline analysis toward real-time online classification
📈 Impact
- Practical ability to deploy deep-learning models on production-floor hardware, matching 'smart manufacturing' and industrial AI adoption needs
- Understands the bandwidth / latency / reliability / thermal trade-offs of edge inference and can evaluate it against cloud architectures