Advisor: Professor Ravi NetravaliDeveloped Gemel, a system for edge-based video analytics that uses novel memory-management techniques to share weights across computer vision models and reduce workload memory costs and swapping delays (NSDI 2023).Developed Boggart, a system for accelerating retrospective machine learning-based video analytics queries via generalizable ingest-time processing (NSDI 2023).In Privid, explored the tradeoffs between privacy and utility in video analytics on public surveillance cameras, and developed a system for privacy-preserving video analytics based on formal differential privacy techniques (NSDI 2022).Studied the adverse effects of delay-based congestion control on HTTP; observed that delay-based algorithms often underestimate available network capacity in mobile networks which enables HTTP/1.1’s multiple connections to achieve higher levels of network throughput than with HTTP/2’s single multiplexed connection (CoNEXT 2020).* I spent the first two years at UCLA; I transferred with my advisor to Princeton University in 2021.

Advisors: Dr. Victor Bahl, Dr. Mani Kotaru and Dr. Francis Yan5G+ML

Mentors: Professor Scott Shenker, James McCauley[Packet State Load Balancing]Explored the concept of packet state load balancing using Transport Control Protocol headers; investigated security benefits of applying cryptography to stateless load balancingContributed implementations of the Hasty Pudding Cipher in Python and P4 (https://github.com/neilsagarwal/hpc)[Hapax]Explored new link state routing mechanism to handle failure more gracefully and reduce the number of packets lost while the routing table is being updated. The idea centers around using reliable flooding to handle packet delivery during recalculation of the routing table

Mentor: Professor Sylvia RatnasamyCourse: CS268 Computer Networks (Graduate)Designed a congestion control protocol based on erasure coding, network enforced fairness, and priority queueing.

Mentors: Professor Raluca Ada Popa, K. ShankariCourse: CS261 Computer Security (Graduate)Designed protocol for distributed differential privacy for aggregation of time-series mobile sensing data.Incorporated encryption for long term secure storage and hardware enclaves to provide the security that existing schemes depend on third party trusted authorities for.

Mentor: Dr. Nathan DeBardelebenDesigned health models of the lab’s High Performance Computing (HPC) clusters used for nuclear simulations and weapons testingDeveloped SaNSA, a software infrastructure that consumes data from multiple sources across the lab such as system logs, compute job reservations, and the system workload manager. It then aggregates the data using a multi-pass state extraction algorithm, and performs a panel of analytics such as anomaly detection via machine learning and state transition analysis using Markov Chains.Used SaNSA to perform a study of 4 open compute clusters at LANL, ingesting and analyzing over 1.1 billion lines of system logs. Detected that a low-performing set of nodes were all part of the same rack – an issue and correlation not noticed by the monitoring team.

Mentors: Professor John Kubiatowicz, Nitesh MorProject: The Global Data Plane (Sponsored by the Semiconductor Research Corporation)Investigated a new architecture for IoT infrastructure, raising the level of abstraction to a data-centric design focused around the distribution, preservation and protection of informationDesigned and developed a distributed application to monitor and visualize the Global Data Plane

Designed a personalized real-time televised sports recommendation system for PlayStation VueDeveloped system in Java using Spring, Docker, Kafka, Cassandra, and Solr

TA for CS61A: The Structure and Interpretation of Computer ProgramsLed daily discussion and lab sections, produced course content and material, assisted in writing and grading exams, and held weekly office hoursGuest lectured to 400+ students on interpreter design, covering concepts such as lexical/syntactic analysis and meta-circular evaluationLecture Slides: https://inst.eecs.berkeley.edu/~cs61a/su16/assets/slides/22-Interpreters_II_full.pdf

Responsible for managing and overseeing logistics for all Indus events and activities including Bollywood Berkeley, a college dance competition

Taught Algebra and Computer Science (Java) classes to underserved 7th/8th gradersDesigned 6-week curriculum (Lesson plans, lecture slides, homework, cumulative project)Professional Development: Lesson planning, classroom management, curriculum design