I am a first year PhD student at the University of Oregon (UO) advised by Prof. Suyash Gupta. I am studying Federated Learning and privacy-preserving techniques in federated learning, especially using Differential Privacy. Before joining UO, I worked at Cisco as a Production Engineer on the Webex team, and completed a research fellowship at ISRO (Indian Space Research Organisation).
I am a PhD student in Computer Science at the University of Oregon, working in the Distopia Lab. My research focuses on privacy-preserving federated learning and machine unlearning, with emphasis on developing computationally efficient methods for federated unlearning.
My academic journey has taken me through Manipal University (Bachelor's) and Stevens Institute of Technology (Master's) before joining University of Oregon for my PhD. I bring a unique combination of industry experience from Cisco and Chubb Insurance, along with research experience at ISRO and MNIT.
Machine unlearning—the ability to remove the influence of specific data from a trained model—is crucial for privacy compliance (GDPR's "right to be forgotten"). However, in federated learning settings, unlearning becomes particularly challenging due to the distributed nature of data and computation.
Caffeine addresses this challenge by developing computationally efficient methods for federated unlearning that avoid costly complete model retraining while maintaining model performance and providing formal privacy guarantees.
Developed a multimodal RAG-based chatbot using large language models for ISRO technical documentation.
Built real-time anomaly detection models for video surveillance systems using computer vision.
Developed NLP models for analyzing customer sentiment in cell phone reviews.
Created ML models to predict employee turnover with interpretable insights for HR.
Middleware 2025 - Doctoral Symposium
This work presents Caffeine, a novel approach to federated unlearning that addresses the computational challenges of removing data influence from distributed machine learning models without complete retraining.
ECIR 2025 - European Conference on Information Retrieval
TheoremView introduces a novel approach to mathematical document search and retrieval, combining traditional information retrieval with mathematical formula understanding.
ISAI 2025 - International Symposium on Artificial Intelligence
MERIT presents a unified framework for learning multimodal embeddings that excel at both retrieval and downstream information tasks across text, image, and structured data modalities.
AusDM 2025 - Australasian Data Mining Conference Submitted
DIVERGEMENT introduces innovative methods for analyzing divergent data patterns in complex datasets with applications in anomaly detection and distribution shift analysis.
Working on privacy-preserving federated learning and machine unlearning. Published paper at Middleware 2025 Doctoral Symposium on the Caffeine project.
Teaching Introduction to Software Engineering. Topics: Git/GitHub, Docker, CI/CD, Agile methodologies, and software testing.
Developed multimodal RAG-based chatbot using large language models for technical documentation and knowledge management.
Maintained production infrastructure for Webex services. Implemented monitoring systems, optimized microservices, and improved system reliability.
Built machine learning models for employee attrition prediction and insurance risk assessment. Conducted data analytics for business insights.
Conducting lab sessions and mentoring students on modern software development practices. Topics include version control, containerization, CI/CD pipelines, and Agile methodologies.
Department of Computer Science
University of Oregon
Eugene, OR 97403
I'm always open to discussing research collaborations, industry projects, or speaking opportunities related to privacy-preserving machine learning and federated learning.
Send Email