November 4, 11, 18 and 25, 2026
About this Workshop
This comprehensive four session workshop dives deep into the engineering techniques required to deploy AI models on resource constrained Edge devices for critical industrial applications. We move beyond academic theory to tackle real world challenges like processing high speed sensor data, predictive maintenance, and radar signals, where microsecond latency and power efficiency are mandatory.
While these optimization strategies can be applied across any platform, understanding the specific target hardware provides a massive advantage during implementation. For instance, we will explore why structured pruning yields vastly different performance results compared to unstructured pruning when targeting FPGAs and specialized accelerators.
This is a fully hands on experience. Each week you will be presented with practical challenges to apply what you have learned directly to real world scenarios, ensuring you leave with concrete implementation skills.
What You Will Learn
- Compression Techniques: Master practical implementations of quantization, pruning, knowledge distillation, and low rank factorization.
- Hardware Optimization: Understand how to align neural network architectures with Edge hardware constraints for maximum inference speed and efficiency.
- End to End Deployment: Gain hands on experience taking optimized models from software directly to target hardware using the KalEdge platform.
Schedule
November 4, 11, 18 and 25, 2026. Four sessions of 2 hours each. (Time to be announced). Sessions will be held online.
- Session 1 (Intro): Why is it important to compress models for Edge AI? Understanding Optimization vs Compression. What are the main hardware challenges? Which key metrics should we track?
- Session 2 (Quantization): Deep dive into FP32 to INT8 conversion, sub byte precision, and fixed point representations.
- Session 3 (Pruning): Exploring unstructured and structured pruning techniques and their impact on hardware architectures.
- Session 4 (Advanced Methods): Knowledge distillation, low rank factorization, and practical deployment using KalEdge.
Pricing and Details
- Capacity: Limited to 10 participants for a highly interactive experience. Once capacity is reached, registrations will close and a second cohort may be evaluated.
- Early Bird: 500 Euros excluding VAT (Valid until October 18, 2026).
- Standard Pricing: 600 Euros excluding VAT (After October 18, 2026).
- Student Scholarships: 2 limited spots available for active students at a special rate of 300 Euros VAT included (subject to review and proof of enrollment).
- Certification: A certificate of participation will be awarded to all attendees upon completion of the workshop.
Note: Applicable VAT will be calculated and added during the checkout process.