Get in Touch

Course Outline

Overview of Edge AI and Nano Banana

  • Defining the key attributes of edge AI workloads
  • Examining Nano Banana's architecture and core capabilities
  • Contrasting edge-based versus cloud deployment strategies

Readying Models for Edge Environments

  • Selecting models and conducting baseline assessments
  • Addressing dependency and compatibility factors
  • Exporting models for subsequent optimization steps

Techniques for Model Compression

  • Exploring pruning methods and structural sparsity
  • Applying weight sharing and parameter minimization
  • Assessing the effects of compression

Quantization for Enhanced Edge Performance

  • Methods for post-training quantization
  • Workflows for quantization-aware training
  • Implementing INT8, FP16, and mixed-precision strategies

Accelerating Performance with Nano Banana

  • Leveraging Nano Banana acceleration features
  • Integrating ONNX and specific hardware backends
  • Benchmarking the performance of accelerated inference

Deploying to Edge Hardware

  • Embedding models into mobile or embedded applications
  • Configuring runtime settings and monitoring systems
  • Resolving common deployment challenges

Profiling Performance and Analyzing Trade-offs

  • Managing latency, throughput, and thermal limits
  • Balancing accuracy against performance metrics
  • Employing iterative optimization techniques

Best Practices for Sustaining Edge AI Systems

  • Managing versioning and continuous updates
  • Handling model rollbacks and compatibility issues
  • Addressing security and system integrity

Recap and Future Directions

Requirements

  • A solid grasp of machine learning pipelines
  • Hands-on experience with Python-based model development
  • Working knowledge of neural network architectures

Intended Audience

  • ML engineers
  • Data scientists
  • MLOps practitioners
 14 Hours

Number of participants


Price per participant

Testimonials (1)

Provisional Upcoming Courses (Require 5+ participants)

Related Categories