Bio:
Vivek Natarajan is a Research Scientist at Google leading research at the intersection of large language models (LLMs) and biomedicine. In particular, Vivek is the lead researcher behind Med-PaLM and Med-PaLM 2, which were the first AI systems to obtain passing and expert level scores on US Medical License exam questions respectively. Med-PaLM was recently published in Nature and has been featured in The Scientific American, Wall Street Journal, The Economist, STAT News, CNBC, Forbes, New Scientist among others. More recently, Vivek also led the development of Med-PaLM M, the first demonstration of a generalist biomedical AI system.
Over the years, Vivek’s research has been published in well-regarded journals and conferences like Nature, Nature Medicine, Nature Biomedical Engineering, JMLR, CVPR, ICCV and NeurIPS. It also forms the basis for several regulated medical device products under clinical trials at Google, including the NHS AI award winning breast cancer detection system Mammo Reader and the skin condition classification system DermAssist.
Prior to Google, Vivek worked on multimodal assistant systems at Facebook AI Research and published award winning research, was granted multiple patents and deployed AI models to products at scale with hundreds of millions of users.
Vivek is also an active angel investor and has invested in a handful of companies working in AI, health, longevity and climate change. Vivek also serves as an AI and technical advisor to several startups and nonprofits.
CV
Google Scholar
Google Research
Contact:
I am always happy to discuss AI and its application in healthcare, bio & robotics among others.
Please reach out to me on Twitter @vivnat or use calendly.com/natviv to set up a chat :)
Recent news:
Sept 2023 - Profile with Analytics India Magazine - Meet the Genius behind Med-PaLM 2
Sept 2023 - Upcoming seminar at Stanford Biomedical Engineering department
Sept 2023 - Talk at ApplySci Boston, MIT on How LLMs can help us scale world class healthcare to everyone
August 2023 - Keynote at IIT Madras with Krishnamurthy Dvijotham on Recent Advances in Multimodal Medical AI at Google.
August 2023 - Appearance on the Cognitive Revolution Podcast with Tao Tu on Med-PaLM M
July 2023- Med-PaLM M on arxiv
July 2023 - Med-PaLM published in Nature with bloomberg article on the backstory
July 2023 - Seminars at Computational & Systems Immunology Seminar Series, Stanford and Brookings Institution.
June 2023 - Talk on Med-PaLM at the Research and Applied AI Summit (RAAIS), 2023 in London, UK
May 2023 - Appearance of the NEJM AI Grand Rounds podcast with Dr Andrew Beam and Dr Arjun Manrai (Harvard University) with my teammate, Dr Alan Karthi (Google)
May 2023 - Appearance on The Harry Glorikian Show with my teammate, Dr Shek Azizi
May 2023 - Appearance on The Cognitive Revolution Podcast talking about recent work on Med-PaLM
May 2023 - Lecture video for Biomedical Transformers as part of CS 25 Stanford course now online
May 2023 - Our paper Med-PaLM 2 now out on arxiv with expert level performance on medical question answering. Med-PaLM 2 was featured in Sundar Pichai’s Google I/O keynote with a promo video on Youtube.
May 2023 - Panel on Large Language Models in Healthcare at SAIL 2023, Puerto Rico, with Zak Kohane (Harvard DBMI), Sebastian Bubeck (Microsoft Research) and Belwadi Srikant (Suki AI).
April 2023 - Talk at Stanford MedAI Seminar on Foundation Models for Medical AI
April 2023 - Our paper, Robust and Efficient Medical Imaging with Self-Supervision has been accepted for publication at Nature Biomedical Engineering. We are also pleased to announce Medical AI Research Foundations, a repository of open source foundation medical AI models in collaboration with Physionet.
April 2023 - Talk and panel on Generative AI in Healthcare at the 6th Illinois Health Data Analytics Summit
March 2023 - Our new state of the art medical Large Language Model, Med-PaLM 2, announced at Google Health Check Up. Med-PaLM 2 is the first AI system to reach expert level performance on MedQA USMLE dataset. Articles in MedPage Today, Scientific American and The Economist.
Feb 2023 - Talk on Transformers in Biomedicine at Stanford CS25 Transformers United course
Feb 2023 - Appearance on the New England Journal of Medicine AI Grand Rounds podcast with Harvard professors Dr Andy Beam and Dr Arjun Manrai and my colleague Dr Alan Karthikesalingam. Episode out in May.
Feb 2023 - Talk on Large Language Models in Medicine at BrainX community event [Video]
Feb 2023 - Talk on AI in Medicine at Imperial College, London
Feb 2023 - Quoted on the potential of ChatGPT replacing jobs
Jan 2023 - Appearance on the Pioneer Park podcast talking about my journey in AI research and Medical AI in particular.
Dec 2022 - Our work on Large language models encode clinical knowledge is now online. Our models reach state of the art on MedQA USMLE with an accuracy of 67.6% exceeding prior work by over 17%.
- Twitter thread
- New Scientist article
- MedPage Today article
- Advisory Board article
- Dr Eric Topol’s blog
- Dr Ethan Mollick (Professor, Wharton University) tweet
- Josh Wolf (GP Lux Capital) tweet
Nov 2022 - Our work on Enhancing the reliability and accuracy of AI-enabled diagnosis via complementarity-driven deferral to clinicians (CoDoC) is now online.
Nov 2022 - Our work on Maintaining fairness across distribution shift: do we have viable solutions for real-world applications? to be presented at NeurIPS
Jun 2022 - Our work on Robust and Efficient Medical Imaging using Self-Supervision is now on arxiv.
Recent talks:
- Transformers in biomedicine Lecture talk - Feb 21, 2023
- Title - Transformers in biomedicine
- Summary - In recent years, the field of AI has been revolutionized by the emergence of Transformers and Large Language Models. However, perhaps nowhere is their impact more profound than in the biomedical domain, where they have the potential to drive human health and potential forward. In this talk, I will explore the ways in which these models are being adapted and applied to biomedical research, with a specific focus on the challenges and opportunities presented by data, modeling innovation, and evaluation. From medical question answering to gene sequencing correction and expression prediction, as well as the natural language annotation of proteins, these cutting-edge techniques are opening up exciting new possibilities for the advancement of medical science and healthcare. Finally, I will offer my insights on how the field of biomedical AI is likely to evolve in the years ahead, and how we can all contribute to this promising area of research.
- Slides
- AI LA life sciences summit 2022 Keynote talk - Oct 20, 2022
- Title - AI in healthcare
- Summary - AI is poised to have a huge impact in healthcare in the coming decade. In this talk, I will provide a view from the trenches on the state of the field; the promise, the challenges and opportunities ahead as we attempt to translate Medical AI from code to clinic. In particular, I will be illustrating this through some seminal works in industry and academia covering medical imaging, medical records, genomics, life sciences among others. I will conclude by laying out a vision for how we can leverage recent advances in Foundation Medical AI to accelerate this translation.
- Slides
- Video
- Deep Learning Classics and Trends seminar - Oct 14, 2022
- Title - Self Supervised Learning for Medical Imaging
- Summary - In this talk, I will introduce some of the last mile translational challenges in taking Medical AI from code to clinic. I will then illustrate why self-supervised learning might be a key piece of this puzzle and cover some of my team's recent works in this space. Finally, time permitting, I will sketch out how the field of medical AI will likely evolve in the coming years and hopefully excite you all to contribute to this space!
- Slides
- Curai External Speaker Series - May 18, 2022
- Title - Taking Medical AI from code to real world impact
- Summary - There has been incredible progress in Medical AI with models often reaching expert level accuracy or uncovering novel insights in diverse applications such as medical imaging, EHR modeling and genomics among others. However, there continues to exist several translational challenges which have prevented broad adoption of medical AI in clinics till date. In this talk, I will introduce some of these translational challenges particularly from an ML perspective and illustrate how we have gone about solving them in taking medical AI from code to clinic in the context of the DermAssist application at Google Health (an AI tool for detecting skin conditions from smartphone images)
- Slides
- AI in Medicine - March 11, 2022
- Title - Taking Medical AI from code to real world impact
- Summary - This talk is an overview of the translational challenges of medical AI illustrated with a case study on DermAssist. I then proceed to lay out the short term opportunities for AI particularly in the post pandemic world as well as a long term vision for foundational medical AI to truly enable precision medicine at scale.
- Slides
- UCL Clinical AI journal club - July 26, 2021
- Title - Self Supervised Learning for Medical Imaging
- Summary - This talk is an overview on why I think self supervised learning is a key solution to some of the grand translational challenges in medical AI and covers some of my team’s recent work in this space.
- Slides
- Video
- On Deck Health talk - May 28, 2021
- Title - AI in Healthcare
- Summary:
- AI is poised to have a huge impact in healthcare in the coming decade. In this talk, I will provide a view from the trenches of the state of the field, its promise and the potential and the challenges we need to overcome to translate AI from research to real world clinical utility and patient impact at scale.
- Other speakers included Vineeta Agarwala (a16z), Malay Gandhi (Rock Health) and Adam Goulburn (Lux Capital) among many other experts and investors.
- Slides
- ReWork AI in Healthcare and Pharma Virtual Summit - March 26, 2021
- Title - Building Better Medical AI for Clinical Deployment at Scale
- Summary:
- In recent years, we have seen several research breakthroughs demonstrating the potential of AI in healthcare settings. However, we are yet to see AI have any impact in the real world and improve patient outcomes. In this discussion, I will lay out some of the key challenges of developing and deploying AI at scale in clinical settings and introduce some of my work done at Google towards addressing them. We will then have an open discussion on how we can accelerate the solving of these challenges and realize the potential of AI in clinical settings. Key takeaways
- *Why is AI yet to have real world patient impact? What are the key technical and non-technical challenges we need to address for this to happen?
- *How can we address those challenges systematically? I will be drawing upon examples from my work at Google to illustrate this
- *We have all the key ingredients to address these issues and if we can make systematic progress, we can very soon realize patient impact at scale with AI
- Slides (Coming soon)
- AI in Medicine - MedTech University College, London - Feb 6, 2021
- Title - Building Better Medical AI for Clinical Deployment at Scale
- Other speakers in the event included experts like Dr Eric Topol and Dr Pearse Keane
- Video
- Vizwiz Grand Challenge Workshop, ECCV, Munich - Sept, 2018
- Title - Pythia for Vizwiz, Winner of the Vizwiz Grand Challenge
- Slides
Press:
Here is a non-exhaustive summary of the press my present and past projects have received.
Self Supervised Learning research at Google AI
- Google AI blog: Self Supervised Learning Advances Medical Image Classification
- TechTalks.com: Self Supervised Learning is a game changer for Medical Imaging
Dermatology research at Google AI
- The Verge: Google announces health tool to identify skin conditions
- Fast Company: Google’s new AI dermatologist can help you figure out what that mole is
- BBC: Google AI tool can help patients identify skin conditions
- Bloomberg TV link
- CNET: Google will now help you identify that suspicious mole or rash
- STAT News: Google debuts an AI-powered app to analyze skin conditions
- Engadget: Google's latest AI tool claims to identify common skin conditions
- Mashable: Google's new A.I. health tool turns your phone camera into a dermatologist
- Financial Times: Google launches AI health tool for skin conditions
- The Australian (AU): How Dr Google will check your skin, hair
- Becker’s Hospital Review: Google tool to identify skin conditions aims to combat frequent misdiagnosis by users
- MobiHealthNews: Google gives a sneak peek into its new AI dermatology assistant
- Modern Healthcare: Google to release new AI tool to identify skin conditions
- MedCity News: Google unveils tool for checking skin conditions
- Fortune: Google just revealed 3 new features. Here’s everything you need to know
- USA Today: Use Google services? These five updates will make your lives easier
Underspecification in Modern Machine Learning Research at Google AI
- MIT Tech Review: The way we train AI is fundamentally flawed
- Discover magazine: Google reveals major hidden weakness in machine learning
Multimodal Research at Facebook AI
Conversational AI Research at Facebook AI
- Automatic video captioning: Techcrunch
- Messenger Assistant M suggestions: Techcrunch, Digital trends
- DeepText: Techcrunch