In this talk, Dr. Yadav will be talking about his research on Edge Artificial Intelligence (Edge AI).
Edge Artificial Intelligence (Edge AI) is ushering in a new paradigm where computation moves closer to data sources, enabling real-time perception, privacy-preserving analytics, and autonomous decision-making in the physical world. Central to this evolution is context-aware computing, where systems sense, interpret, and respond to environmental, behavioral, and temporal cues. From gesture-driven interfaces to health monitoring, robotics, and smart infrastructure, tomorrow’s intelligent edge systems must dynamically adapt to changing context rather than operate as static inference engines. However, selecting optimal hardware platforms for these applications remains a non-trivial challenge, given diverse constraints in compute, memory, power, latency, and privacy. To address this, we introduce Edge-X, a cost-factor evaluation workflow that enables systematic selection and deployment of AI models on edge platforms (Conley,Yadav et al., 2025). Edge-X formalizes a step-by-step method to profile resource needs, map workload characteristics, and rapidly deploy applications across heterogeneous devices. Demonstrated through gesture recognition on three hardware tiers, Edge-X provides actionable insights into balancing capability, energy efficiency, and integration complexity.
This talk extends the Edge-X foundation to explore the future of context-aware edge intelligence – where devices evolve from inference units to modelers, adaptive learners, and federated participants capable of on-device personalization and collaborative learning across distributed ecosystems. We examine emerging processor architectures, context-aware model design, low-power continual-learning approaches, and federated protocols that respect privacy while enabling collective intelligence. The presentation will deliver a grounded deployment framework and a forward-looking roadmap toward self-training, situationally-aware, and cooperative Edge AI systems powering next-generation interactive, intelligent environments.
