Every year, Georgia Tech’s Center for Research into Novel Computing Hierarchies (CRNCH) selects a cohort of Ph.D. student projects that exemplify the cross-stack mission of the center.  

Three students - Vaibhav Bhosale, Vima Gupta, and Nikola Maruszewski - were selected for the CRNCH 2026 Ph.D. Fellowship based on their research proposals. The new fellows will present their work at the annual CRNCH Summit in February 2027.  

Bhosale is a computer science Ph.D. candidate advised by Professor Ada Gavrilovska and Senior Research Engineer Ketan Bhardwaj. His research focuses on distributed systems for Low Earth Orbit (LEO) compute clouds. His work uses the predictable paths of satellites to build a control system that keeps applications running as satellites move. 

Earth observation satellites have onboard GPUs, but onboard energy limits what the model can do. Bhosale’s proposal for the fellowship, Karna: Energy-Adaptive Inference Serving in Space, treats reconfiguration as a control action and schedules it based on the orbital forecast, aiming to serve more queries in orbit without drawing more from the battery than it can safely supply. 

“I look forward to sharing my work with, and learning from, the interdisciplinary community of researchers and faculty at Georgia Tech that CRNCH brings together,” Bhosale said.  

Gupta is a fourth-year Ph.D. student in computer science advised by Assistant Professor Anand Iyer. She is a member of the NEXS Lab. Gupta’s research centers on building efficient artificial intelligence (AI) systems that adapt to both workloads and the underlying hardware. She is currently focused on improving the energy and resource efficiency of agentic AI systems. 

Her project, PICKET: Harness–Hardware Co-Design for Efficient Agentic Inference on Heterogeneous Systems, bridges the communication gap between agentic AI software and hardware. This would allow the system to make smarter decisions about compute and memory, improving energy efficiency without hindering performance.  

“Ideally, I would like my work to help the next generation of AI systems to be more hardware-informed and for the hardware to be harness-aware,” Gupta said. “Personally, I would like this fellowship to shape my Ph.D. thesis and research directions.” 

Maruszewski is a computer science Ph.D. student advised by Associate Professor Josiah Hester. In his research, Maruszewski builds software and hardware systems that make small, power-limited devices more capable.  

Batteryless sensors collect and send all readings back to a base station, wasting limited energy on sending readings that aren’t useful. Maruszewski’s proposal, Classifying With Random Bits: MTJ p-Bits as an Inference Accelerator for Batteryless Sensors, lets the sensor flag which samples might be useful and send only those.  

“This fellowship gives me a semester to take this idea from theory to a working design,” Maruszewski said.