Resource-Constrained Computer Vision

Developed lightweight computer vision and machine learning solutions for agricultural, industrial, and public safety applications, including peanut maturity classification using hyperspectral feature extraction, chlorophyll estimation in winter wheat, smartphone-based color matching for smart textiles, keyway angle detection in manufacturing systems, and vision-based social distancing measurement for autonomous vehicles.

In-Memory / In-Sensor Computer Vision

Developed a post–von Neumann visual system integrating sensing, memory, and computation using memristive crossbar arrays. By pushing dot-product operations and classification directly into hardware with analog WTA and mixed-signal peripherals, the system enables low-power, real-time edge vision.

Hardware-Aware Learning / Neuromorphic Computing

Designed energy-efficient spiking neural networks using current-controlled memristive synapses and hardware-aware mapping. The framework accounts for device variability, ADC limits, and non-idealities, while incorporating sleep-inspired reservoir computing for adaptive, memory-constrained learning.