Nicholas Soures

Lead systems engineer focused on AI/ML integration.

Leading research on efficient temporal learning and continual learning algorithms, focusing on scalable solutions for real-time and embedded AI applications. Spearheading an NSF-funded project on adaptive neural architectures for lifelong learning in dynamic environments. Mentoring high school, undergraduate, and graduate students in advanced AI techniques and neuromorphic computing. Developing collaborative frameworks for cross-disciplinary teams, integrating AI models into hardware-constrained systems. Contributing to grant proposals and technical reports for new research initiatives in embedded and neuromorphic AI.

Designed and deployed embedded AI systems on FPGA and ASIC platforms, enabling low-power, low-latency performance for SWaP-constrained defense applications. Developed physics-informed generative models to simulate complex data environments, improving model robustness in real-world deployment scenarios. Co-developed photonic and neuromorphic accelerators for high-speed matrix operations and AI inference, pioneering novel hardware-software integration methods. Collaborated across hardware, software, and product teams to transition innovative ML algorithms from prototype to field-ready solutions. Contributed to proposals and technical roadmaps for next-generation embedded AI technologies.

Ph.D. Thesis: Lifelong Learning in Spiking Networks Through Neural Plasticity - Development of energy efficient, task agnostic lifelong Pioneered energy-efficient, task-agnostic lifelong learning frameworks for spiking neural networks (SNNs), achieving up to 20% performance gains over state-of-the-art methods. Co-developed digital and analog AI accelerators for on-device learning with metaplasticity, funded by AFRL. Led ML system development for activity recognition on event-based cameras, improving temporal resolution for NSA-sponsored research. Collaborated with UT Health and Southwest Research Institute to model COVID-19 spread using SEIR and LSTM models, providing actionable insights via a public dashboard. Authored and secured ~$2M in grant funding, leading interdisciplinary research efforts across AI and neuromorphic systems. Mentored graduate students and helped design a “Brain-Inspired Computing” course, fostering the next generation of neuromorphic AI talent.

Achieved state-of-the-art performance in video activity recognition by developing multi-layer neural networks with 10x reductions in computation and memory requirements, enabling deployment on embedded platforms. Explored random CNN architectures for energy-efficient image processing, providing innovative solutions for SWaP-constrained environments. Conceptualized and simulated neuromorphic hardware systems (digital, analog, and photonic) to accelerate AI algorithms under NSF and AFRL funding. Evaluated Intel TrueNorth neuromorphic chips for low-power image recognition, advancing understanding of hardware-constrained AI performance. Contributed to hierarchical classification systems using transfer learning and autoencoders for NSA-sponsored projects.

Courses Taken: Digital Systems Design 2, Digital Signals Processing, Analytical Topics, Brain inspired computing, Computer Vision, Digital IC design, Intro to Principles of Statistics and Data Mining, Pattern Recognition, Deep Learning, Engineering Analysis

Courses Taken: Digital Systems Design 2, Digital Signals Processing, Analytical Topics, Brain inspired computing, Computer Vision, Digital IC design, Intro to Principles of Statistics and Data Mining, Pattern Recognition, Deep Learning, Engineering Analysis

Relevant Courses: Electronics 1, Mathematical Methods in Electrical Engineering, Power Systems Analysis, Intro to Python, Quantum Mechanics, Electro-Magnetic Theory, Thermo-dynamics, Optics, Classical Mechanics, Security, Discrete Math, Linear Algebra, Calculus 1-3, Ordinary Differential Equations, Partial Differential Equations