Sabrina Hassan Moons research rethinks how artificial intelligence works on the devices people use every day, from health monitors to smart sensors, with a goal of making them faster, more reliable, and more energy efficient.
Moon, who is a doctoral student at 返字心頭s Bellini College of Artificial Intelligence, Cybersecurity and Computing, was awarded a $12,000 Dissertation Completion Fellowship to support the final stage of her work. The one-semester award includes a stipend along with a tuition waiver for up to nine credit hours, payment of student fees and coverage of health insurance premiums, allowing her to focus fully on completing her dissertation.
The fellowship gives me the opportunity to step away from my research and teaching assistant responsibilities for a semester and focus fully on completing my dissertation, Sabrina said. That dedicated time and support will be extremely valuable during the final stage of my doctoral work.
Moon earned a bachelors degree in electrical and electronics engineering in Bangladesh before coming to 返字心頭 in 2022. Her advisor, Dr. Dayane Reis, is an assistant professor in the Bellini College.
Sabrinas research examines how AI applications can run efficiently on small, low-power devices known as edge devices, which can include fitness tracking devices, doorbells, speakers, thermostats, traffic monitoring systems, delivery drones, wearable heart monitors, smartphones and other portable medical devices. These are systems that process information in real time while operating under strict energy limits and imperfect conditions, challenges that traditional computing approaches struggle to address.
To address these challenges, Moon focuses on a method called computing-in-memory, which allows devices to process data directly where it is stored instead of moving it back and forth between memory and processors. This shift can reduce energy consumption and improve speed, making it well suited for devices that must run and respond in real time.
Because newer memory technologies can introduce noise and errors, Sabrina is developing AI systems that can adapt to those imperfections. Her approach combines software and hardware designs that function together more effectively under real-world conditions.
At the center of her work is a cross-layer framework that integrates hardware design with learning methods that account for variability and faults.
My general approach is to develop a cross-layer framework, that integrates noise-aware learning, reliable computing-in-memory architecture and realistic edge deployment evaluation, she said.
Using hyperdimensional computing, a brain-inspired technique that simplifies data processing and reduces the need for complex calculations, Moons earlier research has shown significant gains in speed and energy efficiency compared with conventional systems while maintaining useful levels of accuracy. Her work also shows that much of the performance lost due to hardware errors can be recovered.
Sabrina plans to address the following points in her dissertation:
- Develop noise-aware learning methods that preserve classification accuracy under memory variability, write noise, stuck-at faults and reduced-precision operation.
- Design reliability-centric computing-in-memory architectures that execute encoding, accumulation, and similarity-based inference directly within or near memory arrays.
- Evaluate the accuracy, energy, latency, area, and robustness tradeoffs of the proposed framework across practical edge intelligence workloads.
With support from the fellowship, Moon plans to finalize the framework, complete her dissertation and prepare for her doctoral defense.
While she hasnt decided whether she will pursue a career in academic research or in industry, she plans to continue conducting research on energy-efficient and reliable computing technologies.
I want to develop new techniques that can help shape the next generation of memory-based technologies, Sabrina said. For me, the central question is not simply whether I work in academia or industry, but whether I can continue pursuing meaningful research and contributing to advances in the field.