Going Beyond Sensors Everywhere
ABSTRACT
Sensors are everywhere, in our homes, cities, factories, vehicles, and wearables, and yet we’re only scratching the surface of their potential. What happens when every object can sense, process, and act? How do we rethink the boundaries of system design?
This talk explores the evolution toward smart sensors, devices that not only measure the world but also decide when and how to respond, like a nervous system for our connected environments. At the core of this shift is embedding AI/ML directly at the sensor edge, enabling real-time, local decision-making and reducing the need for constant communication with gateways or the cloud.
We’ll explore architectural choices such as phased wake-up, where higher-power components remain asleep until intelligent triggers activate them, and discuss how quantized, low-footprint AI/ML models are deployed on constrained microcontrollers. We’ll also address how synthetic data helps overcome the challenge of scarce labeled datasets for edge training and validation enabling fast time to market for the systems in question TDK is at the forefront of this transformation, developing sensors and modules with embedded intelligence that drive efficient system behavior. But challenges remain: How do we optimize what stays at the edge and what moves to the cloud? What are the obstacles to scalable deployment? How do we architect systems for energy, latency, privacy, and maintainability? Join us for a technical and strategic view into the emerging world of smart sensing, and how it’s reshaping the way we build and interact with the intelligent systems of tomorrow.
BIOGRAPHY
Juan S. Mejia Santamaria is an engineering leader with over 16 years of experience driving innovation across machine learning, sensors, embedded systems, signal processing, and control systems. Currently serving as a Principal Engineer and Team Lead at TDK USA Corporation, Juan leads a team developing AI/ML algorithms and synthetic data pipelines for IMU-based classification and anomaly detection in industrial IoT settings.
He has a proven track record in architecting efficient, low-power ML solutions for edge devices, including smart wearables and embedded sensor platforms. Notably, he led the market release of TDK’s Sensor Inference Framework (SIF), enabling decision tree deployment to sensors with ultra-constrained memory and compute resources, and contributed to the development of ML2MCU, a rapid model deployment tool for quantized DNNs.
Prior to TDK, Juan held engineering roles at Cruise and Western Digital, where he scaled teams, drove embedded validation processes, and led control algorithm design for high-volume manufacturing of HDDs. Juan holds a Ph.D. in Systems and Entrepreneurial Engineering from the University of Illinois at Urbana-Champaign, where his research focused on model predictive control for autonomous vehicles. He is passionate about building robust software/hardware systems, enabling intelligent sensing, and leveraging AI/ML to accelerate innovation at the edge."