Sweta Priyadarshi

About Sweta Priyadarshi

Currently, I am focused on enhancing the efficiency of generative AI technologies, such as Large Language Models, Large Vision Models and Large Multimodal System, by optimizing their accuracy for edge devices with stringent computational and memory constraints. My role at Qualcomm AI Research involved pioneering hardware-aware model optimizations and contributing to multiple patent applications that significantly improve the intersection of AI model efficiency and hardware capability.I am deeply committed to exploring the intricacies of deep neural network architectures and their operational efficiencies, particularly in the realm of knowledge distillation, quantization aware training, Post training quantization, pruning, LoRA based finetuning, and Neural Architecture Search. This focus stems from my extensive 4+ years of experience in both academia and industry, where I have developed novel deep learning algorithms that address real-world challenges in AI deployment.With a strong foundation in Machine Learning from Carnegie Mellon University and over four years of hands-on experience in cutting-edge AI research, I continually strive to push the boundaries of technology to achieve remarkable improvements in the field of AI. For a detailed look at my scholarly contributions, including papers presented at international conferences like ICCV, ICML and ICLR, please visit my GitHub page: https://github.com/Swetap24 or Google Scholar Page & hl=en

Experience

1. Research Scientist

   NVIDIA

   Jul 2024 — Present · Santa Clara, CA, US

Education

• Carnegie Mellon University

  Master of Science - MS, Electrical and Computer Engineering

  2019 — 2020

• Manipal Institute of Technology

  Bachelor of Technology, Electrical and Electronics Engineering

  2014 — 2018