Ajay Karthikeyan

LLMs, RecSys, Ads@DoorDash | GenAI@Amazon Science | ML@Microsoft Research

United States

About

Experience

  • Machine Learning Engineer at DoorDash
    Jul 2024 - Present · 2 yrs

    Embedding training efforts for search ranking and retrieval.

  • Thesis at University of Illinois Urbana-Champaign
    Jan 2024 - Apr 2024 · 4 mos

    LLMs for Code

  • Applied Scientist Intern at Amazon
    May 2023 - Aug 2023 · 4 mos

    Gen AI at Amazon Ads: Image Editing using Diffusion Models Responsible for the first image editing effort in the team using text-to-image diffusion models - Implemented mask-augmented image editing and fine-tuned these models (initialized from Stable Diffusion) to improve edit performance by 13%. - Implemented zero-shot image editing by manipulating internals of the diffusion denoising process.

  • Microsoft Research (India)
    • Research Fellow
      Jul 2019 - Feb 2022 · 2 yrs 8 mos

      Team: Machine Learning & Optimization, Applied Science and Engineering - Research and implementation of a quantized neural network-based decision tree learning algorithm that enables end-to-end gradient-based training of trees. [Paper]: https://openreview.net/pdf?id=u0n1chY0b6 - Worked on a zeroth-order optimization based algorithm that improves performance of software systems by replacing handwritten heuristics with data-driven ones [Paper]: https://www.microsoft.com/en-us/research/publication/selftune-tuning-cluster-managers/

    • Research Intern
      Jan 2019 - Jun 2019 · 6 mos