Graduate student at UCSD, focusing on Distributed Systems

GPA: 3.88/4.00

Relevant Courses:

• Distributed Systems
• Deep Learning Systems
• LLM System Optimization

Projects:

• System Measurement: Developed a suite of micro-benchmarks to evaluate the Rockchip RK3588S SoC, utilizing ARMv8 cycle counters to measure CPU scheduling and OS primitive latencies with nanosecond precision.(Link)
• Distributed Container Snapshotting: Built a Chandy-Lamport based snapshotting system with a reliable-UDP sidecar proxy to checkpoint and restore live, distributed containerized applications.(Link)
• Modern GPU DSLs: Benchmarked RMSNorm and Flash Attention-2 across Helion, Triton, Gluon, and CuTe on four NVIDIA GPU generations to evaluate performance/portability trade-offs.(Link)

GPA: 3.96/4.00

Achievements:

• Academic scholarship for securing 2nd rank out of 60+ students

Relevant Courses:

• Operating Systems
• Computer Networks
• Cloud Computing
• Big Data Analytics

Projects:

• BitTorrent Client: Built a peer-to-peer file-sharing client using Python’s AsyncIO and BitTorrent protocol with a custom Bencode parser
• GIST: Developed a YouTube video summarizer using NLTK, BART model, SQLite, and Tkinter

• Profiling the attention operator in transformer LLM inference across prefill (prompt/KV-cache lengths) and decode (batch size/KV-cache lengths), fitting analytical models to measured compute and communication latencies
• Building a trace-driven simulator over multi-turn LLM conversation traces to compare parallelism strategies (CP/TP/DP + EP) for a disaggregated prefill/decode serving deployment

• Continued voluntary collaboration with the DREAM:Lab to finalize research validation for the JPDC 2025 publication
• Authored the in-depth system architecture and scalability analysis for Flotilla, leading to its acceptance in the Journal of Parallel and Distributed Computing (JPDC)
• Orchestrated containerized deployments using custom Docker images across distributed clusters to validate system performance on 1000+ concurrent clients

Responsibilities include:

• Built an asynchronous federated learning framework (Flotilla) in Python, optimized for edge hardware deployment
• Implemented server and client sides using MQTT and gRPC for efficient message passing and coordination in federated learning
• Designed a custom Redis-based state store with checkpointing to enable recovery from full server failures without data loss or disruption
• Integrated client selection and aggregation strategies from current research for performance, accuracy, and turnaround optimization
• Collaborated with PhD students under Prof. Manik Gupta (BITS Pilani) and Prof. Yogesh Simmhan (IISc) to ensure Flotilla’s scalability and reliability
• Configured and managed an 80+ node edge cluster (Nvidia Jetsons, Raspberry Pis), supporting lab infrastructure and projects including Flotilla

Volunteering:

• Senior Student Volunteer, Indian Institute of Science – IEEE/ACM CCGrid 2023: Co-organized a 300+ participant conference; coordinated 3 poster sessions and assisted keynote speakers and faculty
• Student Volunteer, IISc Open Day 2023: Coordinated presentation sessions for DREAM:Lab projects

Responsibilities include:

• Taught a graduate-level course to a class of 40+ students, comprising topics such as HDFS, Map-Reduce, Apache Spark
• Facilitated and led a 2-hour lab session per week, prepared and graded assignments, conducted one-on-one office hours, and conducted doubt-clearing sessions

Responsibilities include:

• Developed Intellza, a unified data storage and analytics platform, alongside a cross-functional team
• Developed and integrated a module to maintain and track schema changes for MongoDB on Intellza using LiquiBase
• Created Docker images and optimized the existing images as per Docker’s recommendations, reducing the image size to 35% and improving the build times of the project’s CI/CD pipeline by 50%