Pramod Chunduri

I’m a PhD Candidate in Computer Science at Georgia Tech advised by Prof. Joy Arulraj, where I’m building efficient systems for AI applications, with a particular emphasis on video analytics and large language models (LLM).

My research focuses on accelerating query processing on AI platforms with a goal of achieving low latency and high accuracy. Through my research, I have developed advanced video analytics systems capable of answering complex queries involving multiple objects, actions, and motions in videos. Presently, I am optimizing Retrieval-Augmented Generation (RAG), a state-of-the-art LLM paradigm, to enhance question-answering efficiency.

In addition to my academic pursuits, I have gained practical industry experience as a Research Scientist Intern at Adobe’s real-time experiences (REAL) lab. There, I utilized machine learning techniques to devise personalized video editing tools capable of recommending optimal highlights and edits from raw footage.

news

Feb 29, 2024	SketchQL is accepted at SIGMOD’25! SketchQL introduces a novel sketch-based interface to query videos efficiently.
Feb 15, 2024	I will be interning at AWS AIRE as an Applied Scientist Intern in summer’24.
Dec 3, 2023	Actively seeking internship opportunities in efficient LLM and video analytics pipelines. Contact me!
Oct 28, 2023	Exciting news! Our GitHub repository exploring advanced LLM-based RAG pipelines has been trending on Hacker News!
Sep 30, 2023	Check out the stargazers-reloaded application to analyze your favorite GitHub communities using the power of LLMs.
Apr 18, 2023	Seiden is accepted at VLDB’23! Seiden builds an efficient query-agnostic index to accelerate video analytics.
Aug 15, 2022	Concluded a rewarding internship at Adobe Research on personalized video editing using ML.
Jun 20, 2022	Presented my Video Analytics paper Zeus on efficient action localization at SIGMOD’22.

selected publications

2025

SIGMOD

SketchQL: Video Moment Querying with a Visual Query Interface

Renzhi Wu*, Pramod Chunduri*, Ali Payani, Xu Chu, Joy Arulraj, and Kexin Rong

SIGMOD, 2025

2022

SIGMOD

Zeus: Efficiently localizing actions in videos using reinforcement learning

Pramod Chunduri, Jaeho Bang, Yao Lu, and Joy Arulraj

In SIGMOD, 2022

Abs PDF Code

Detection and localization of actions in videos is an important problem in practice. State-of-the-art video analytics systems are unable to efficiently and effectively answer such action queries because actions often involve a complex interaction between objects and are spread across a sequence of frames; detecting and localizing them requires computationally expensive deep neural networks. It is also important to consider the entire sequence of frames to answer the query effectively. In this paper, we present ZEUS, a video analytics system tailored for answering action queries. We present a novel technique for efficiently answering these queries using deep reinforcement learning. ZEUS trains a reinforcement learning agent that learns to adaptively modify the input video segments that are subsequently sent to an action classification network. The agent alters the input segments along three dimensions - sampling rate, segment length, and resolution. To meet the user-specified accuracy target, ZEUS’s query optimizer trains the agent based on an accuracy-aware, aggregate reward function. Evaluation on three diverse video datasets shows that ZEUS outperforms state-of-the-art frame- and window-based filtering techniques by up to 22.1x and 4.7x, respectively. It also consistently meets the user-specified accuracy target across all queries.