I'm Huynh "Lucius" Vo — a Ph.D. candidate at Oklahoma State University (OSU) working at the seam where neuromorphic hardware, in-memory computing, artificial intelligence, and optimization meet.
Begin readingFrom growing up in Vietnam, to training as an engineer in Finland, and now conducting research in the US, each chapter of my journey has equipped me with a new discipline. I carry forward the patience of optical metrology, the scale of semiconductor manufacturing, and the complex, open questions of software-hardware co-design. This site is an exhibition of that path and a window into what I am building next: in-memory computing systems that no longer separate memory from thought.
— L. V., Stillwater, MMXXVIFrom a childhood in Vietnam to research benches in Finland, factory floors at Intel, and now the prairies of Oklahoma — each place reshaped the questions I wanted to ask.
A childhood between books and circuit boards
My journey did not start in a university laboratory, but in Vietnam, driven by a childhood fascination with natural philosophy. Long before I was designing computational systems, I was trying to understand the basic rules of physics and chemistry. The questions that drive my research today took root there—quietly, and before I even knew the math to describe them.
B.Tech., Electrical and Electronics Engineering
My first real engineering apprenticeship. It began with a cross-border indoor sensor collaboration between Metropolia University of Applied Sciences and Germany's Osnabrück Hochschule, and culminated in a bachelor's thesis on wireless excavator safety. I taught myself the patience that hardware demands, and earned Tutor of the Year for the company I kept among incoming students.
M.Sc., Automation and Electrical Engineering — Data Science
At the Metrology Research Institute in Aalto University, my research scope rapidly expanded. Working alongside Dr. Petri Kärhä and Dr. Erkki Ikonen, I investigated the differential spectral responsivity of solar cells, and modeled Maximum Power Point Tracking (MPPT) for photovoltaic (PV) panels under the mentorship of Dr. Amir Sepehr. Beyond energy systems, I broadened my focus to author a pioneering Life Cycle Assessment of the Samsung Galaxy Smartwatch. But quietly, running parallel to the hardware and modeling, I fell into machine learning during my master's research—and never came out. I graduated with honors, bridging the physical and the algorithmic with a thesis on deep learning for PV module fault diagnosis.
Manufacturing and Deep Learning Engineer
When COVID-19 paused my PhD prospects at VTT (Technical Research Centre of Finland), I returned home, unaware that this setback would become my greatest catalyst. Stepping into the industry, I had the chance to deploy my research at massive scale. At First Solar, I developed defect-commonality tooling that remains in active use across their factories. Moving to Intel, I fully transitioned into an industrial researcher, leading the Intelligent Inspection and Automatic Disposition (ILLIAD) system—highlighted at the 2023 AI Everywhere Conference—and teaching Python and ML across the global organization.
Ph.D., Industrial Engineering and Management
My current home. I work on resistive random-access memory (RRAM), in-memory computing, and the optimization algorithms that can take advantage of both. Two-time finalist in the Institute of Industrial and Systems Engineers (IISE) Student Paper Competition, Society for Industrial and Applied Mathematics (SIAM) Travel Awardee for PP26, and currently President of the OSU Institute for Operations Research and the Management Sciences (INFORMS) Chapter.
Ph.D. Research Student Aide
A summer at a US national laboratory, asking whether RRAM crossbars can solve the linear programs that move electricity across the U.S. grid. The preliminary answer became a paper accepted at SIAM PP26.
My research lives at the meeting of three disciplines. Resistive random-access memory (RRAM) gives us a substrate where computation and storage are no longer separate things. Analog and in-memory computing turns that substrate into a working machine. Optimization and learning ask what such a machine could solve that GPUs cannot.
I am building the bridge between the three — designing distributed primal-dual algorithms that run natively on RRAM crossbars, and chasing the day when neuromorphic hardware quietly outperforms a rack of GPUs on the world's hardest linear programs.
Anne Vallayer-Coster — The Attributes of Music, 1770
A glimpse at what currently occupies my desk. The full list lives in publications.
We present a distributed primal-dual hybrid gradient (PDHG) method co-designed for resistive random-access memory (RRAM) in-memory computing, enabling large-sc…
We introduce MELISO+: a distributed, full-stack RRAM in-memory computing framework with integrated two-tier error correction for scalable, high-dimensional mat…
We present SplitVAE, a decentralized scenario generation framework that leverages variational autoencoders to generate high-quality scenarios without moving st…
We introduce MELISO (In-Memory Linear Solver), a comprehensive end-to-end VMM benchmarking framework tailored for RRAM-based systems. MELISO evaluates the erro…
Always glad to meet new collaborators — for research, mentoring, or simply a conversation about David's brushwork over coffee.