About Me
| Senior Staff Software Engineer | Intel Corporation |
PhD Student
School of Computer Science
Embedded Pervasive Lab
Advised by Umakishore Ramachandran
Georgia Institute of Technology
Advised by Professor Umakishore Ramachandran
Research Direction
My research is centered on the systems-level challenges of building safe, reliable, and scalable edge computing platforms for real-time autonomous applications, with a primary focus on connected autonomous vehicles. My approach is grounded in the philosophy that before such systems can be deployed in the complex and dynamic real world, we must first solve for foundational correctness and then explicitly engineer for scalability.
My research strategy, therefore, proceeds in two main thrusts.
Foundational Reliability
First, I focus on establishing foundational reliability by identifying and mitigating failure modes that arise from the interaction between distributed components. This involves moving beyond viewing factors like network latency as performance bottlenecks and instead treating them as potential sources of critical state-consistency errors that can compromise application correctness. The goal here is to design and validate lightweight, protocol-level solutions that guarantee a consistent and trustworthy worldview, providing a stable base layer upon which more complex logic can be built.
Engineering for Scalability
Second, with a reliable foundation established, my work addresses the challenge of computational scalability. As we move from sparse scenarios to dense, dynamic environments, the finite compute resources of the edge node itself become the primary bottleneck. My research here explores the design of adaptive edge services that are aware of their own resource constraints. By utilizing mechanisms for graceful degradation—such as dynamically managing input load, adjusting processing fidelity, and prioritizing safety-critical tasks, the system can make intelligent, real-time trade-offs. This ensures that core safety and performance guarantees are maintained even when operating under heavy computational pressure.