Brain-Inspired Intelligence for Smart Sensors and Edge Devices
Imagine devices that can learn continuously from their environment, answer questions about what they sense, and protect your privacy—all without sending your data to the cloud. We design brain-inspired computing systems using Hyperdimensional Computing (HDC), a computational paradigm that mimics how the human brain encodes and processes information. HDC enables ultra-efficient machine learning on resource-constrained devices like smartwatches, home sensors, and environmental monitors. Our research focuses on lifelong learning algorithms that adapt to new patterns without forgetting old knowledge, and hardware-software co-design that makes these systems practical for real-world deployment on FPGAs and edge processors.
Teaching Sensors to Understand and Communicate
What if your smart home could answer questions like "Did I leave the stove on this morning?" or "When was the last time someone watered the plants?" We build multimodal sensor systems that combine cameras, microphones, motion sensors, and environmental monitors to create a comprehensive understanding of daily activities. Using large language models and neuro-symbolic reasoning, our systems can interpret long-term sensor data and respond to natural language queries. This research bridges IoT sensing, natural language processing, and human-computer interaction to make ambient intelligence truly useful and accessible.
Autonomous Systems for Agriculture and Environmental Monitoring
Modern agriculture faces critical challenges: labor shortages, climate variability, and the need for sustainable practices. We develop intelligent robotic systems for tasks like crop monitoring and harvesting, combining drones, ground robots, and sensor networks into collaborative teams. Our research addresses real-world deployment challenges—operating in unpredictable outdoor environments, coordinating multiple autonomous agents with limited communication, and making split-second decisions with limited computational resources. From precision agriculture in smallholder farms to large-scale environmental monitoring, we build systems that work reliably in the field, not just in the lab.
Privacy-Preserving Collaborative Intelligence for IoT Networks
What if hundreds of smart devices—from wearable health monitors to home sensors to agricultural robots—could learn together without sharing your private data? We design distributed learning systems where devices collaboratively improve their intelligence while keeping data local. Our frameworks handle the messy reality of IoT deployments: devices with different capabilities, intermittent connectivity, and the need to adapt quickly to changing conditions. This approach enables applications ranging from personalized healthcare monitoring to coordinated swarms of agricultural robots, all while protecting user privacy and reducing reliance on cloud infrastructure.
