Isha Garg

About Isha Garg

I am a Deep Learning Research Engineer at Apple with 13+ years of experience, ranging across academia (PhD at Purdue), research engineering positions (2.5 years at Apple; 2 years at NUS), and hardware engineering (4 years at Synopsys and AMD)My current focus is on designing efficient, reliable and scalable algorithms for large vision and language models. Before that, I focused on designing efficient and scalable memory compilers.My work has been featured in many premiere machine learning conferences; including a Best Paper Award at symposium, Spotlight/Oral talks at ICML and ICLR, and other works published in CVPR, ICCV and Nature. For more details, please refer to my Google Scholar: & hl=en

Experience

1. Deep Learning Research Engineer

   Apple

   Mar 2023 — Present · Seattle, WA, US

   Goal: Improve the efficiency and quality of large vision and language models• Designed an algorithm to decrease the number of parameters in Vision Transformers by 4×- Best Paper Award at GENAI4Edge at Spring Symposium 2025 - Presented at Efficient Systems for Foundation Models Workshop, ICML 2024 - Featured on Apple Machine Learning Blog at https://machinelearning.apple.com/research/revealing-utilized-rank - Leveraging insights from this to improve the internally shipped models• Architected a change in the LLM shipping pipeline to improve quantization error- This reduced the model shipping time from two weeks to two days, impacting the entire model lifecycle timeline.Current Focus:• Improved downstream task adapters: Working on better initializations for LoRA adapters for large models, and their unification across downstream tasks• Connecting Sparsity and Low-Rank Structure: Designed an algorithm to convert identified low-rank subspaces into sparsity in large language models• Personalization and Long Context: Designed an algorithm to efficiently distill user context, with the ultimate goal of enabling personalization

Education

• Purdue University

  Doctor of Philosophy - PhD, ECE

  2017

• Birla Institute of Technology and Science, Pilani

  Bachelor of Engineering (B.E.), Electronics and Instrumentation Engineering

  2009 — 2013

Skills

• Circuit Design
• Digital Electronics
• Vlsi
• Spice
• Saliency
• Python
• Microprocessors
• Microcontrollers
• Labview
• Vhdl
• Fpga
• Modelsim
• Neuromorphic Computing
• Matlab
• Electronics
• Cadence
• Unix
• Sram
• Cadence Virtuoso
• Digital Circuit Design
• Deep Learning
• Caffe
• Ltspice
• Machine Learning
• Semiconductors
• Analog Circuit Design