Blake A. Wilson, Ph.D.
- Interim CEO, Zephram — inference optimization & multi-agent systems - Ph.D. Electrical & Computer Engineering, Purdue University - Expertise: CUDA optimization, ML-driven inverse design, RISC-V SoC software, quantum algorithm synthesis - Industry: ARM, QuEra Computing, Quantinuum
Research interests: ML for physical systems; high-performance computing (CUDA); quantum circuit synthesis; photonics inverse design.
Machine learning for physical systems — CUDA, inverse design, and quantum-adjacent ML
I am a computer engineer often working in algorithmic optimization and machine learning, with a focus on physics-informed AI, inverse-design and hardware-software co-design. My research is deeply-interdisciplinary, involving collaborations with physicists, engineers, and mathematicians from institutions such as Oxford, Cambridge, and Harvard to develop optimized software and models for complex scientific problems, having published in venues such as Applied Physics Reviews, Nature Partner Journal, American Controls Conference, CLEO, and APS. I have led cross-functional efforts at the intersection of AI and hardware—including founding the NanoML team and serving as PGA team lead at the Quantum Science Center while developing ML-driven quantum circuit synthesis at QuEra. My industry experience spans research roles at Purdue SoCET, QuEra, and Quantinuum, and I currently serve as interim CEO of Zephram, where I focus on full-stack multi-agent systems and inference optimization for science at large. My proprietary work while at Quantinuum includes collaborations with NVIDIA and DeepMind-aligned teams for advancing RL and transformers for quantum circuit synthesis. My work has been supported by NASA Ames, AFRL, NSF, DOE, ORNL (OLCF), and AWS Braket. To date, I have mentored 200+ researchers through programs such as SURF, SoCET alongside my personal capacity. My students have won numerous best first-time researcher awards and poster awards at conferences and internal symposiums.
Collaborators & Network
Institution logos: University of Oxford, University of Cambridge, Harvard University, University of California, Berkeley, Purdue University, Oak Ridge National Laboratory, Sandia National Laboratories, NVIDIA, Microsoft.
Industry and proprietary R&D
Summarized for confidentiality
Industry ML for Quantum Computing and Quantum Chemistry
Led ML research and engineering for RL training of GPT2-scale models to optimize quantum circuits with applications in quantum chemistry. Developed proprietary multi-GPU CUDAQ kernels out-performing in-house chemistry calculation speed and developed data-collection pipeline for molecular data.
Inference optimization and multi-agent systems platform (Zephram)
As interim CEO, I lead a team of 8 Oxford PhDs and engineers developing a multi-agent RL platform and its underlying infrastructure. We have major proprietary breakthroughs in context compaction, finetuning, and sampling.
Research Outcomes
Leadership
DOE National Lab Summer School & PGA
Led the graduate and postdoc association for a DOE national lab consortium at Oak Ridge. Organized and moderated panels across three annual summer schools on ML for science, helping raise $200k+ in funding and bringing together researchers from national labs and universities.
AFTx04 System-on-Chip -- Purdue SoCET
Led the software team on Purdue's SoCET (System on Chip Engineering Team) building the AFTx04 RISC-V processor. Designed and programmed the ROM and bootloader for initiating processor verification and debugging. Built a Python-to-RISC-V transpiler for generating Verilog ROM and benchmarked on FPGAs.
Researchers mentored
Mentored 200+ researchers through programs such as SURF, SoCET. My students have won numerous best first-time researcher awards and poster awards at conferences and internal symposiums.
Selected publications
Deep learning in photonic device development: nuances and opportunities
Iyer, V., Wilson, B.A., Chen, Y., Kildishev, A.V., Shalaev, V.M., Boltasseva, A.
npj Nanophotonics 3, 5 (2026)
Machine-learning-assisted photonic device development: a multiscale approach from theory to characterization
Chen, Y., Montes McNeil, A., Park, T., Wilson, B., Iyer, V., Bezick, M., Choi, J., Ojha, R., Mahendran, P., Singh, D.K., Chitturi, G., Chen, P., Do, T., Kildishev, A., Shalaev, V., Moebius, M., Cai, W., Liu, Y., Boltasseva, A.
Nanophotonics, Vol. 14, Issue 23 (July 2025)
Machine Learning Framework for Quantum Sampling of Highly-Constrained, Continuous Optimization Problems
Wilson, B., Kudyshev, Z., Kildishev, A., Shalaev, V., Kais, S., Boltasseva, A.
Applied Physics Reviews, 8, 041418, (Impact Factor: 19.16)
In the news
Coverage from SPIE, Hackster.io, Purdue Engineering, and others.