CV

Masters

University of Virginia, Charlottesville, VA

Computer Science

• Privacy in large language models
• Inference attacks in machine learning models
• Differential Privacy

Bachelors in Engineering (B.E.)

UIET, Panjab University, Chandigarh, India

Computer Science

Researcher

Openmined

• Conducted research on the disparate impact of DP-SGD and PATE on minority groups in the dataset.
• Conducted research on wide neural networks and how they are able to perform better than deep neural net- works under specific conditions like differential privacy.

Software Developer Intern

OmniVision Technologies, Santa Clara, CA

• Deployed and debugged using GDB industry-standard automated testing methods like Crest and KLEE for C and C++ codebase
• Migrated these methods to Python, reducing software dependencies which come along with C++
• Developed a novel instrument-free method for automated unit test generation which reduced processing time by 10 times and is currently being used by the algorithm team internally

Research Intern

Oracle Labs, USA

• Lead research in the privacy group focusing on prevention of training data leaks in large language models (LLMs)
• Fine-tuned LLMs using HuggingFace DeepSpeed and techniques such as LoRA in Pytorch
• Patented a new concept of entity-relationship privacy in LLMs which targets a more realistic definition of privacy at the entity-level instead of measuring verbatim training data outputs

Research Intern

Samsung Research Institute, Noida, India

• Lead research on improving the existing time series-based human activity recognition algorithms in Galaxy watch5 and contributed 10K lines of code to the codebase
• Fine-tuned generative adversarial networks in Pytorch and reduced noise in the videos by 5 times compared to traditional denoising filters in Samsung smartphones