Aakash Kaku

I'm a Senior Software Engineer at Google DeepMind and a Ph.D. graduate from the NYU Center for Data Science at the Courant Institute of Mathematical Sciences. My research and engineering focus lies in developing cutting-edge machine learning systems, particularly in healthcare and multilingual LLMs.

Previously, I contributed to Bard, Google Assistant, and Gemini LLMs. I also led research at NYU on stroke rehabilitation and brain segmentation.

Contact: ark576[at]nyu.edu. Google Scholar | Resume

Education

  • New York University - Center for Data Science (Courant)

    Ph.D. in Data Science (Sept 2019 – Dec 2023)
  • New York University - Center for Data Science (Courant)

    MS in Data Science (Sept 2017 – May 2019)
  • Indian Institute of Management Bangalore

    PGDM (equivalent to MBA), Top 5% (Jun 2014 – Mar 2016)
  • Institute of Chemical Technology, Mumbai

    B.Tech in Chemical Engineering, Top 10% (Aug 2010 – May 2014)

Research / Publications

  • Gemini: A Family of Multimodal Models – Contributed to Google's flagship multimodal system [ArXiv]
  • Stroke Rehabilitation with Wearables – Developed ML methods to quantify rehab for stroke patients [Nature npj Digital Medicine]
  • Brain Segmentation and Cartilage Detection – Pioneered techniques in MR segmentation at NYU [GitHub]
  • Intermediate Layers in Contrastive SSL – Published at NeurIPS 2021 [ArXiv]
  • COVID-19 Prediction in ER – Part of team featured in Nature npj Digital Medicine
Complete List

Teaching

  • Mathematical Tools for Data Science – NYU CDS (TA & Grader, Spring 2021)
  • Deep Learning for Medicine – NYU Med School (TA, Spring 2019)
  • Quantitative Methods 2 – IIM Bangalore (TA, 2015)

Work Experience

  • Senior Software Engineer, Google DeepMind

    Nov 2024 – Present
    • Led Gemini for education visual learning initiative; boosted student engagement with novel visual learning tools.
    • Built a multilingual tool-routing system; increased API success and reduced TPU costs.
    • Developed RAG-based multilingual query generation for Search to mitigate LLM hallucinations.
  • Software Engineer, Google

    Dec 2023 – Nov 2024
    • Advanced RLHF post-training for non-English LLMs (RLXF, IPO methods).
    • Built scalable reward model pipelines and auto-eval systems for international LLM deployment.
  • Research Intern, Google Research

    Jun 2023 – Aug 2023
    • Improved multilingual reward models (Bard) and semantic parsing (Assistant).
  • Intern, Philips Research NA

    May 2022 – Aug 2022
    • Detected acoustic shadows in ultrasound using weak supervision segmentation models; reduced radiologist burden by 85%.
  • Strategy Analyst, Accenture Consulting

    May 2016 – Aug 2017
    • Saved $3M for a Fortune 100 client with a custom NLP automation tool.
    • Advised chemical sector M&A execution during due diligence.