I am a third year PhD student at Georgia Tech, started in August 2022.
Previously, I was an undergrad at Penn State, graduated in May 2022.

PhD advisors: Prof. Ada Gavrilovska and Prof. Kexin Rong

Research interest

Currently, I am working on reducing latency in AI/LLM inference systems for emerging workloads.
Broadly, my research interests lie in making practical systems for real-world use.
Past research:
Built a profiling tool, Lotus [1] [2], to uncover bottlenecks causing poor GPU utilization in ML preprocessing pipelines.

[1] Lotus: Characterization of Machine Learning Preprocessing Pipelines via Framework and Hardware Profiling (IISWC '24)
Problem: Preprocessing has become a bottleneck in ML training pipelines leading to poor GPU utilization.
Solution: Introduced Lotus, a lightweight profiling tool that:

  • Captures fine-grained events and maps them to low-level CPU metrics.
  • Diagnoses variability in per-batch preprocessing time (e.g., poor core provisioning, GPU stalls from OOO batch arrivals).
  • Identifies microarchitectural bottlenecks — e.g., front-end stalls in CPU-based preprocessing as core counts increase.

[2] Lotus: Characterize Architecture-Level CPU-Based Preprocessing in ML Pipelines (HotInfra '24)
Demonstrates that Lotus helps infrastructure designers evaluate CPU SKU choices for ML training servers.

Etc

    GitHub: @rajveerb
    Linkedin: @rajveermb
    X: @rajveerbach

Contact


Electronic: rr [at] gatech [dot] edu (click here)
Physical: Klaus Lab 3319, 266 Ferst Dr NW, Atlanta, GA 30332