Edge AI & Optimization Lab

Overview

Not every AI model runs in a data center. The Edge AI Lab teaches you to take models that were trained on powerful GPUs and compress, optimize, and deploy them on resource-constrained edge devices. The lab has two components: a cloud-based T4 GPU environment for running TensorRT optimizations, ONNX conversions, and quantization experiments, plus remote SSH/VNC access to physical NVIDIA Jetson Orin boards and Coral Edge TPU devices at our Houston training center. You'll measure real inference latency, power consumption, and accuracy on actual hardware — not just simulated benchmarks.

What You'll Do in This Lab

• Convert PyTorch and TensorFlow models to ONNX intermediate format
• Optimize models with TensorRT — FP16, INT8 calibration, layer fusion
• Quantize models for TensorFlow Lite and Coral Edge TPU deployment
• Deploy optimized models to physical NVIDIA Jetson Orin via remote access
• Benchmark inference: measure FPS, latency, power consumption, and accuracy on real hardware
• Build complete edge pipelines — camera input → preprocessing → inference → output

Lab Workflow