Event
Ph.D. Research Proposal Exam: Everest Bloomer
Thursday, November 12, 2026
12:00 p.m.
1146 AVW
Souad Nejjar
301 405 8135
snejjar@umd.edu
ANNOUNCEMENT: Ph.D. Research Proposal Exam
Name: Everest Bloomer
Committee:
Dr. Sahli Shah (Chair)
Dr. Shuvra Bhattacharyya
Dr. Timothy Horiuchi
Date/time: Thursday, November 12th, 2026. 12PM-1:30PM
Location: 1146 AVW
Title: Analog In-Situ Compute to transform NP-Class Problems into Approximate Tractable Domains
Abstract: Modern computation is constrained by the unsustainable power consumption and high execution times required to solve Nondeterministic Polynomial Time (NP) class optimization problems on traditional hardware. To address these concerns, a systematic co-design methodology that leverages analog in-situ computing and Ising machine physics to map NP-class problems into approximate, tractable domains is introduced. First is established a novel, substrate-independent framework that unifies heterogeneous computing paradigms by decomposing computation into three distinct abstraction layers: the Problem Space, the Computational Space, and the Physical Semantics. To demonstrate how this theoretical framework guides physical architecture design, DIRRK is developed, a deterministic differential Ising solver algorithm co-designed with a Resistive Random Access Memory (ReRAM) in-memory computing architecture. DIRRK reformulates the Ising Hamiltonian into a first-order differential system accelerated by ReRAM crossbar arrays, shifting execution from unpredictable stochastic annealing to deterministic, density-dependent convergence. DIRRK is validated using a hardware prototype fabricated in the SkyWater 130nm CMOS process. Finally, to evaluate the framework's practical utility, this analog co-design is applied to optimize the NP-hard Search and Rescue (SaR) scheduling problem for autonomous drone swarms. DIRRK and the SaR problem serve to provide support for both the foundational benchmarking theory and the hardware implementation pathway for the development of new scalable, deterministic, and energy-efficient accelerators.
