Event
Ph.D. Dissertation Defense: Jasvith Raj Basani
Friday, November 20, 2026
9:00 a.m.
KEB 1110
Announcement: Ph.D. Dissertation Defense
Name: Jasvith Raj Basani
Committee:
Professor Edo Waks (Chair)
Professor Saikat Guha
Professor Julius Goldhar
Professor Carlos A. Rios Ocampo
Professor Norbert Linke (Dean's Representative)
Date/time: Friday, November 20th, 2026 at 9:00 AM
Location: KEB 1110
Meeting Link: https://meet.google.com/ qqx-yomc-baj
Title: Scaling Logic and Control in Photonic Quantum Processors
To translate native optical operations into universal, error-correctable quantum logic, we develop quantum optical neural networks that combine machine-learning-based control of multimode interference with engineered nonlinear interactions. This framework unifies the essential primitives of error-corrected computation, including logical state preparation and encoding, universal gate operations, and nondestructive syndrome extraction. Addressing the integration and error-scaling challenges of larger processors, we demonstrate how interferometric mesh topology governs performance and enables effective mitigation of errors arising from component imperfections. Building on this relationship, we develop the generalized Green Machine, which exploits multiplexing techniques to realize programmable interferometric architectures without the quadratic hardware overhead of conventional spatial processors.
The ability to program complex quantum transformations also opens new possibilities for learning unknown physical processes, particularly those whose output statistics are classically intractable. Reproducing such distributions from experimental measurements, however, can demand prohibitive learning resources. Our entanglement-enhanced adversarial framework harnesses shared quantum correlations to establish a provable quantum speedup in sample complexity over separable learning strategies. By structuring the learning objectives, it further establishes stable convergence while mitigating barren plateaus that limit conventional variational learning. Experiments on a programmable nanophotonic processor demonstrate accelerated convergence and a widening reduction in measurement requirements across the tested system sizes, advancing the prospect of learning directly from physical experiments.
Together, these advances establish foundations for scalable photonic quantum computation while extending quantum advantage beyond computation alone. They point toward physical learning machines capable of learning from quantum experiments, reproducing complex physical phenomena, and advancing scientific discovery beyond classical computational limits.
