What is Edge AI?
Edge AI refers to artificial intelligence algorithms that run directly on local devices—smartphones, IoT sensors, microcontrollers, and embedded systems—rather than in the cloud. This approach enables real-time decision-making, enhanced privacy, reduced latency, and lower bandwidth costs.
Why Edge AI Matters
Traditional AI relies on sending data to cloud servers for processing. Edge AI flips this model:
- Privacy: Data stays on the device, never leaving the user’s control
- Speed: Millisecond response times without network round-trips
- Reliability: Works offline and in low-connectivity environments
- Efficiency: Reduces energy consumption and infrastructure costs
Key Concepts
Model Optimization
Running AI on resource-constrained devices requires specialized techniques:
- Quantization: Reducing model precision (e.g., from 32-bit to 8-bit)
- Pruning: Removing unnecessary neural network connections
- Knowledge Distillation: Training smaller models to mimic larger ones
Hardware Platforms
Edge AI runs on diverse hardware:
- Microcontrollers (MCUs): Arduino, ESP32, STM32
- Single-board computers: Raspberry Pi, NVIDIA Jetson
- Neural Processing Units (NPUs): Dedicated AI accelerators
- FPGAs: Programmable hardware for custom workloads
Frameworks & Tools
Popular tools for Edge AI development:
- TensorFlow Lite: Google’s framework for mobile and embedded
- ONNX Runtime: Cross-platform inference engine
- PyTorch Mobile: Meta’s mobile deployment solution
Getting Involved
The EDGE AI FOUNDATION organizes work through specialized groups staffed by Strategic Partners and Academia:
- Datasets and Benchmarks; how do we develop and contribute high quality datasets and models to the community and establish best practices on comparing “performance”?Chair: Adam Fuks, NXP Semiconductors
Read the Working Group white paper
Visit EDGE AI Labs for datasets, community code, challenges and more. - Marketing & Industry; Effective storytelling to highlight the impact of “AI in the real world” across our partner organizations with market research and analyst outreach and bring the voice of end users and industry segments into our community via Blueprints livestream, solution analysis, and industry symposiumsChair: Eric Smiley, embedUR
- Neuromorphic; how do we accelerate efforts like neurobench for comparisons, cooperation with MATRIX and neuro-commons and other key areas for industry and academic benefit?Chair: Dr. Petrut Antoniu Bogdan, Innatera
- Generative Edge AI; addressing the entire stack from metal to cloud – exploring all of the innovations that enable generative and agentic AI on the edge, as well as key scenarios, acting as the Technical Program Committee for our Generative Edge AI online forum series.Co-Chairs: Danilo Pau, STMicroelectronics, Professor Hajar Moussanif, Cadi Ayyad University Morocco
Read the Working Group white paper - Audio; how do we establish best practices in audio AI, canonical scenarios, benchmarks and best integrate academic researchChair: Tomer Badug, Elia Shenberger, Ceva
- Node Learning; Promotes research and collaboration around decentralized intelligence that learns incrementally and continually, both on individual nodes and across heterogeneous connected networked nodesChair: Prof. Eiman Kanjo, Nottingham Trent University / Imperial College London
Co-Chair: Dr. Wael Guibene, Silicon Labs - Career; development programs to connect and inform professionals students with high performance edge AI careersChair: Luke Perrins, 5V
View the EDGE AI Career livestreams and recordings - Commercialization; chartered with complementing existing working groups by mapping commercial tools and solutions to applicable use cases across the edge continuum.Chair: Jason Shepherd, Co-founder & CEO, Atym
Co-chair: Rob Woolley, Senior Principal Technologist at Wind River
Connect with the Community
- Join our Discord server for real-time discussions
- Attend our EDGE AI San Diego 2026
- Follow working group activities and publications
Next Steps
- Explore the Working Groups to find your area of interest
- Review the Glossary for terminology
- Check out Tools & Libraries for hands-on resources
- Read the Standards & Specs for industry guidelines
