cv

Research Collaborator

Massachusetts Institute of Technology

Multi-agent Reinforcement Learning, Policy Loss Function, Vision Language Models

• Working on LLMs for multi-agent reinforcement learning (MARL), focusing on long-horizon planning and task execution in complex, partially observable environments. Formulated new policy loss function to improve multi-agent decision-making, reducing suboptimal actions by 26%
• Developed Vision Language Models (VLMs) for MARL, enabling agents to understand and execute complex tasks using visual inputs and natural language instructions.

Graduate Research Assistant

Electrical and Computer Engineering Department, NYU Tandon School of Engineering

AI for hardware design, LLM based frameworks

• Conducting research on leveraging Large Language Models (LLMs) for hardware Code analysis, including using LLMs as encoders to predict vulnerable code directly from RTL before synthesis.

Member of Technical Staff

Oracle Corporation

Machine Learning, Log Anomaly Detection, Recommendation System, Plugin Engine, Web Development

• Created ML system to identify errors causing build failures, providing diagnosis, resolution with 98.6% accuracy.
• Built log anomaly detection, processing 10TB/day, 94% accuracy, reducing bug-fixing time from 2 days to 1 hour.
• Devised a recommendation system generating 50,000 daily recommendations, automating bug detection and fixing.
• Engineered a plugin engine, handling 1M+ CPU processes daily, improving productivity by 40%.
• Developed web pages with HTML, CSS, JavaScript, and Python APIs, increasing user engagement by 30%

Master of Science

New York University

Computer Engineering

• Intro to Machine Learning
• Advanced Machine Learning
• Deep Learning
• Machine Learning Operations (MLOps)

Bachelor of Technology

National Institute of Technology Warangal

Electronics and Communication Engineering

• Data Structures and Algorithms
• Object-Oriented Programming
• Java

The recipient of the David Chang and Cecilia M. Chang Student Leadership Award

New York University

The purpose of this award is to support NYU Tandon students in attending regional or national conferences, meetings, competitions, or other events, to develop their potential to become leaders in society, industry, and academia.

Top 20 winners of Qualcomm On-Device AI Hackathon

Qualcomm

One of the top 20 winners of the Qualcomm On-Device AI Hackathon, recognized for developing an OS Copilot AI assistant that enhances user productivity and multitasking. The AI assistant uses on-device machine learning to diagonise and fix issues, improving user experience and device performance.

Best Outgoing Student

National Institute of Technology Warangal

Awarded in recognition of my outstanding overall academic and research performance during my undergraduate studies. This honor reflects my excellence in coursework, impactful research contributions, and commitment to academic growth.

Best Allrounder

National Institute of Technology Warangal

Recognized for excellence in academics, research, and extracurricular activities during my undergraduate studies. Along with strong academic and research contributions, I was an active member of the institute's badminton team, balancing both intellectual and athletic pursuits.

Best Student Paper

National Institute of Technology Warangal

Awarded for outstanding research contributions and scholarly impact during my undergraduate studies. This recognition highlights the significance of my research work, its academic rigor, and its contribution to the field.

Best Student Paper

National Institute of Technology Warangal

Awarded for outstanding research contributions and scholarly impact during my undergraduate studies. This recognition highlights the significance of my research work, its academic rigor, and its contribution to the field.

Merit Scholarship

National Institute of Technology Warangal

Awarded for academic excellence, ranking among the top 10 students in the department. This scholarship recognizes consistent high performance and dedication to academic achievement.

Google Research AI Summer School

Google Research

Selected as one of 150 participants globally for the Google Research AI Summer School, an elite program focused on cutting-edge advancements in artificial intelligence and machine learning. Engaged with leading researchers, explored state-of-the-art AI methodologies, and gained hands-on experience in AI research applications.

International Physics Olympiad

IPhO Committee

Qualified for the prefinal round of the International Physics Olympiad (IPhO), ranking among the top students nationally. This selection reflects strong problem-solving skills, deep conceptual understanding of physics, and excellence in competitive examinations.

General Secretary

Electronic Amateur & HAM Club

Served as the General Secretary (President) of the Electronics Amateurs and HAM Club, leading initiatives in electronics, and machine learning. Organized technical workshops, hands-on projects, and knowledge-sharing sessions, fostering innovation and engagement within the community.

PrefixLLM: LLM-aided Prefix Circuit Design

ArXiv

PrefixLLM leverages large language models (LLMs) to synthesize optimized prefix circuits, transforming the task into a structured text generation problem (SPCR). Using an iterative design space exploration (DSE) framework, it achieves 3.70% area reduction while maintaining the same delay, demonstrating the potential of LLMs in arithmetic circuit synthesis.

Detecting anomalies in log messages using code-derived message patterns to guide structured message classification

US Patent

Developed a log anomaly detection system leveraging code-derived message patterns for structured classification. The system processes 10TB of log data daily with 94% accuracy, reducing bug-fixing time from 2 days to 1 hour by enabling early and precise anomaly detection.

ChildrEN SafEty and Rescue (CENSER) System for Trafficked Children from Brothels in India

AAAI

We propose the ChildrEN SafEty and Rescue (CENSER) system used by the Guria non-profit organization to retrieve trafficked children from brothels in India.

Iteratively reweighted minimax-concave penalty minimization for accurate low-rank plus sparse matrix decomposition

IEEE TPAMI

We propose weighted minimax-concave penalty (WMCP) as the nonconvex regularizer and show that it admits an equivalent representation that enables weight adaptation. Similarly, an equivalent representation to the weighted matrix gamma norm (WMGN) enables weight adaptation for the low-rank part.

Performance analysis of weighted low rank model with sparse image histograms for face recognition under lowlevel illumination and occlusion

CONIT

A comparison of the low-rank recovery performance of two Low-rank matrix approximation (LRMA) algorithms-RPCA and WSNM is brought out on occluded human facial images.

Attendance Management System

ICCCI

We propose a novel method of marking attendance using facial recognition. The proposed method uses small and accurate deep supervised network for recognition of faces in a wild classroom scenario. The proposed method is robust to variations in illumination, pose, and expression.

AEVBComm: an intelligent communication system based on beta-VAE

CSI Transactions on ICT

The paper proposes a new CNN based VAE communication system. The VAE (continuous latent space) based communication systems confer unprecedented improvement in the system performance compared to AE (distributed latent space) and other traditional methods.

Jaya Algorithm Based Intelligent Color Reduction

AEOTIT

A color reduction hybrid algorithm is proposed by applying Jaya algorithm for clustering. We examine the act of Jaya algorithm in the pre-clustering stage and K-means in the post-clustering phase.

OS Copilot

Developed an AI assistant using on-device machine learning to enhance user productivity and multitasking, diagnosing and fixing issues to improve user experience and device performance.

• On-Device Machine Learning
• Natural Language Processing
• Multitasking Enhancement
• User Productivity

OptimUX AI

Developed an Agentic AI framework to recommend optimized website UI designs based on UX best practices.

• Python
• Large Language Models
• Retrieval Augmented Generation
• AI Agents

FitGen

Led the development of an AI-powered fitness app offering personalized workout plans using user data and machine learning algorithms, increasing user adherence by 32%

• Genetic Algorithms
• Video Pose Analysis
• Convolutional Neural Networks
• Git

RallyMaster AI

Designed a unique posture correction algorithm using perspective transformation techniques that analyzed player movements; the system processed over 500 training sessions, providing actionable insights for enhanced gameplay

• Computer Vision
• Video Pose Analysis
• Git

3D Scene Reconstruction

Developed accurate 3D maps by implementing three local feature/matcher methods: LoFTR, DISK, and KeyNetAffNetHardNet. This project showcases my expertise in Structure from Motion techniques and advancing to 3D modeling.

• Computer Vision
• Structure from Motion

Log File Summarization

Developing a log file summarization tool using the Drain algorithm to efficiently extract key insights and patterns from complex log data, enhancing system monitoring and troubleshooting.

• Machine Learning
• Natural Language Processing
• Log Analytics