Welcome back David! Dr. David Wu co-founded our podcast back in the height of COVID and is now a radiation oncology resident at Stanford and advisor to an LLM-based startup called Jaide. Since our founding David has been a driving force in communicating the happenings in the medicine and machine learning space and we are excited to talk about his latest efforts in contributing to this area. Jaide uses patient reported outcomes to give physicians access to predictive modeling on disease evolution and eases physician documentation burden using state-of-the-art large language models. 

00:00:00 - Introduction 

00:01:31 - From feeling aimless after undergrad to a Stanford resident and advisor for an AI startup (Jaide)

00:05:31 - The beginnings of Jaide: Using LLMs to document patient outcomes 

00:07:57 -  LLMs in action - the first clinical trial in Brazil

00:11:03 - How do you use Jaide? 

00:12:30 - Could LLMs take away key skills from physicians’ training?

00:15:14 - What if Step 1 was the entry exam to med school? 

00:18:10 - What drew you to Stanford?

00:21:11 - Have you faced criticism/self doubt about pursuing so many things outside of traditional medicine? 

00:23:08 - Why rad onc?

00:26:20 - Best/worst med school and residency experiences - “I’m just gonna become a monk and retire” 

00:29:33 - Will AI replace doctors? AI isn’t even close in this one area.

00:35:06 - The double edged sword of AI

00:38:41 - How did you choose a specialty? Is the impact of AI important? 

00:39:45 - What got you into hip hop/rap? - “Artists make meaning out of suffering” 

00:43:45 - What gives your life meaning?

00:46:55 - What advice do you have for younger people concerned about the impacts of AI on medicine?

Jaide - jaide.care

Twitter - @davidjhwu 

Host: Nathan Keller

Twitter: @NathanKellerX

Linkedin: https://www.linkedin.com/in/nathankeller1/

Producer: Saurin Kantesaria

Linkedin: Saurin Kantesaria

Dr. Steven Hart is a Senior Associate Consultant in AI at Mayo Clinic who has played a key role in shaping genomics and digital pathology with GenomeGPS, Mayo Clinic’s primary DNA sequencing workflow. His groundbreaking contributions have led to advancements in understanding inherited cancer risk and improving digital pathology workflows. With over 100 peer-reviewed publications, Dr. Hart’s innovative algorithms are driving efficiency in genetic predisposition testing, reducing unnecessary procedures, and enhancing precision healthcare.

We had some audio issues for this one which we tried to fix but they're still pretty apparent so apologies for that D:!

00:00:00 - Introduction 

00:01:09 - From a factory worker to a leader in AI and medicine

00:05:11 - Proving people wrong as a motivator

00:06:37 - Crazy factory stories

00:07:38 - Why Mayo Clinic?

00:09:52 - Surprising things about Mayo Clinic

00:11:33 - Is Mayo Clinic’s data high quality?

00:12:55 - How to prepare healthcare for AI (and why AI won’t actually have the biggest impact)

00:20:50 - Democratizing pathology with AI 

00:25:38 - Will AI replace pathologists?

00:29:24 - How do you judge how well an embedding works? 

00:33:22 - Reducing expectations for diagnostic AI usage in healthcare

00:36:46 - How do you keep up with the rapidly evolving pace of AI?

00:38:31 - OpenAI o1 and prompt hacking 

00:41:27 - Are we close to artificial general intelligence? 

00:47:03 - How helpful are regulatory agencies like the FDA with translating AI?

00:49:52 - What makes a good question?

00:53:33 - Favorite parts about living in Rochester, MN

00:55:14 - What gives your life meaning? 

00:58:36 - Advice for young people in uncertain times

Host: Nathan Keller

Twitter: @NathanKellerX

Linkedin: https://www.linkedin.com/in/nathankeller1/

Producer: Saurin Kantesaria

Linkedin: Saurin Kantesaria

Ran Shaul is the chief product officer and co-founder of K Health. With his robust background as a successful founder, Ran has been pivotal in transforming how we approach medical diagnostics and personalized treatment. Under his leadership, K Health has developed innovative AI-driven solutions, including a partnership with Cedars-Sinai and Mayo Clinic. Ran's dedication to improving the patient experience by leveraging technology is reshaping healthcare delivery, making it more efficient and accessible.

Host: Nathan Keller

Twitter: @NathanKellerX

Linkedin: https://www.linkedin.com/in/nathankeller1/

Producer: Saurin Kantesaria

Linkedin: Saurin Kantesaria

00:00 - Introduction 

00:52 - What are 3 patient questions doctors and AI should help answer? 

03:26 - Why does ChatGPT fall short in diagnosing patients? 

07:30 - AI does the tedious stuff so doctors can focus on medicine (K Health’s model) 

09:29 - Combing through 400,000,000 unstructured doctor’s notes 

11:53 - How do you ask the right clinical questions with AI?

15:41 - Putting a clinician in the loop of AI learning

19:21 - “You can have the perfect algorithm…it does not mean it will be used properly in any clinical setting” 

23:27 - The difficulties transitioning from leading a startup to a larger company

26:56 - Telemedicine 2.0 - integrating 24/7 online care with brick and mortar hospitals (Cedars-Sinai Virtual Platform)

31:32 - AI can go further than notes - helping physicians proactively manage patients

39:07 - What gives your life meaning?

42:56 - What advice do you have for young people?

Dr. Robert Dürichen leads the machine learning analytics team at Arcturis Data, a company focused on processing and analyzing large-scale electronic health record (EHR) datasets. His current research uses small and large language models to enrich EHR datasets from unstructured patient notes and improve quality through standardization techniques.

Hosts: Nathan Keller + Madeline Ahern

Twitter: @NathanKell57664 + @maddie_ahern 

Audio/Video Editor + Art: Saurin Kantesaria

Linkedin: Saurin Kantesaria

Intro 0:00

Who is Robert Durichen? 1:29

What is Arcturis? 6:25

How can machine learning speed up clinical trials? 9:43

Typical Arcturis Project 12:06

Progression of Machine Learning 24:45

AI Taking Jobs 29:50

What is Arctex? 33:50

Who works at Arctex? 38:55

Future of Arcturis 40:58

What gives your life meaning? 43:30

Advice for young people on maintaining a work-life balance 44:26

Dr. Nina Kottler is the associate chief medical officer of clinical artificial intelligence and vice president of clinical operations for Radiology Partners, the largest radiology practice in the US, serving over 3,250 hospitals and other healthcare facilities, interpreting over 53 million exams annually. 

Host: David Wu

Twitter: @davidjhwu

Audio Producer: Aaron Schumacher

LinkedIn: Aaron Schumacher 

Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

00:00:58 What brought you to the intersection of medicine and artificial intelligence?

00:07:00 The importance of translating between clinicians and AI engineers

00:12:54 The origins of Radiology Partners 

00:16:40 Dr. Kottler’s start in Teleradiology 

00:21:18 The transition form analog to digital in Radiology 

00:27:35 The current state of Radiology Partners

00:32:00 When did Dr. Kottler become a leader in the AI projects?

00:45:00 AI models that Radiology Partners use 

00:52:00 Fragility, Technological Evaluation and Business evaluation in Radiology AI systems

00:56:10 Dr. Kottler’s thoughts on what the future of AI and Radiology will look like. 

01:00:30 Dr. Kottler’s advice for people in medicine desiring unique paths. 

01:02:45 What brings you joy?

Munjal Shah is the co-founder and CEO of Hippocratic AI, a new startup in Generative AI + Healthcare. Hippocratic is building a safety-focused large language model specifically built for the healthcare industry. 

Host: David Wu

Twitter: @davidjhwu

Audio Producer: Aaron Schumacher

LinkedIn: Aaron Schumacher 

Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

Time Stamps:

00:00:58 What brought you to the intersection of medicine and artificial intelligence?

00:06:20 Overview of the American Healthcare System 

00:08:06 Hippocratic AI and the Adherence Problem within healthcare 

00:14:30 Building an AI Chronic Care Nurse for specific conditions 

00:17:15 AI systems and medical co-morbidities 

00:24:00 The process of building Hippocratic AI

00:32:45 Becoming more efficient than ChatGPT4

00:33:48 Navigating the problem of hallucinations with Hippocratic AI

00:39:30 How close are we to Health General Intelligence (HGI)?

00:45:40 What advice would you give to someone interested in starting their own company?

00:48:20 How did mentorship shape your path?

00:49:40 What brings you joy?

00:52:25 How do you find novel ideas for start-ups?

Dr. Mamdani is a professor, pharmacist, and epidemiologist. He is the Vice President of Data Science and Advanced Analytics at Unity Health Toronto and Director of the University of Toronto Temerty Centre for Artificial Intelligence Research and Education in Medicine (T-CAIREM). Dr. Mamdani’s team bridges advanced analytics including machine learning with clinical and management decision making to improve patient outcomes and hospital efficiency. Dr. Mamdani is also Professor in the Temerty Faculty of Medicine, the Leslie Dan Faculty of Pharmacy, and the Institute of Health Policy, Management and Evaluation of the Dalla Lana School of Public Health at the University of Toronto. He is also a Faculty Affiliate of the Vector Institute. He has published over 500 studies in peer-reviewed journals.

Host: Raeesa Kabir 

Audio Producer: Melanie Bussan

Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

Social Media: Nikhil Kapur

Time Stamps:

0:00 Dr. Mamdani’s Background and Career Path

9:30 Where current data driven medicine strategies fall short and how AI can step in

17:00 How Dr. Mamdani’s work in AI and machine learning began

22:00 Applied Health Research Center and the Ontario Policy Research Network

28:45 The impact of utilizing machine learning and AI at the level of patient care - Chart Watch

35:50 Logistics of Developing and Implementing AI solutions

39:10 Insights Gained - From Purpose to Implementation

43:30 Directing Multiple Projects - Recruitment of AI Team 

47:45 Future Projects: Back to AI Basics 

54:15 Future of AI in Medicine - Fostering trust in AI

57:20 Advice to Younger Self

CardinalKit (now Spezi) is an open-source framework for Digital Health Applications and Research. They were recently featured in the news for releasing HealthGPT, an experimental iOS app that lets you query your health data. Spezi is housed in the Stanford Byers Center for Biodesign and directed by Oliver Aalami, MD with Vishnu Ravi, MD as lead architect. Also joining us on this interview is postdoc Paul Schmiedmayer, PhD.

Spezi provides a suite of tools to build modern, interoperable digital health tools from the ground up, from the app itself to storing and analyzing collected data in the cloud. It is designed to accelerate rapid prototyping of digital health applications by reducing costs by as much as 75% (~$150,000) and timelines by 12 months.

Host: David Wu

Twitter: @davidjhwu 

Audio Producer + Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

Social Media: Nikhil Kapur

Time Stamps:

00:58 - The expertise behind Spezi (CardinalKit)

08:03 - Healthcare has a lack of data standardization + Why you should know about HL7 FHIR

14:13 - How did Spezi (CardinalKit) become what it is today?

18:26 - Drink Spezi!

19:53 - Making code/healthcare data more modular and user-friendly

26:40 - Translating a med student's sensor research to a useable device for kids with cerebral palsy

31:20 - From a $40,000 eczema patch test in clinic to a completely at-home test

35:45 - Using healthGPT to make health data easy to understand for patients (LLM on FHIR)

42:35 - How do you deal with privacy issues?

49:33 - What do you think the future of AI in medicine will look like in 10-20 years?

52:00 - Applications where using only an LLM doesn't always work (a case for hybrid systems)

55:30 - What brings you joy?

58:43 - What makes a successful digital health team?

Dereck Paul, MD is a cofounder and the CEO of Glass Health, an AI-powered medical knowledge management and clinical decision-making platform that helps clinicians provide better patient care. Previously, he was an internal medicine resident at Brigham and Women's Hospital, Harvard Medical School and a medical student at the UCSF School of Medicine.

Host: David Wu

Twitter: @davidjhwu 

Audio Producer + Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

Social Media: Nikhil Kapur

Time Stamps:

01:13 - From music major to med school to making a startup

06:30 - Poor healthcare technology = physician burnout, the motivation for building Glass Health

09:15 - Glass Notebook - "Notion for doctors"

11:24 - Building a startup in the era of Chat-GPT

13:50 - What doctors need in an AI-assisted diagnosis software

19:15 - Transition towards a more AI oriented technology - Glass AI

23:00 - How does Glass AI make accurate diagnoses?

28:40 - Why doctors need to be involved in building clinical AI products

30:50 - Practical usage of Glass AI in the clinic

33:04 - Why Glass AI will be more trustworthy than Chat-GPT in writing clinical notes

37:43 - Why LLMs don't need to be perfect for use in the clinic

40:28 - Ethical implications of Glass AI and similar products

45:34 - Should we disclose when we use AI to write a clinical note?

49:13 - What do you think the future of AI in medicine will look like in 10-20 years?

52:30 - What brings you joy? What gives your life meaning?

56:10 - Would you ever go back to being a musician?

Jerry Liu is the co-founder and creator of LlamaIndex (formerly known as GPT-Index), an interface that allows users to connect their data to LLM’s such as Chat-GPT. He has a B.S. in Computer Science from Princeton and has worked at companies such as Quora, Uber, and Robust Intelligence prior to starting LlamaIndex.

Host: David Wu

Twitter: @davidjhwu

Audio Producer: Aaron Schumacher

LinkedIn: Aaron Schumacher 

Video Editor + Art: Saurin Kantesaria

Instagram: saorange314

Social Media: Nikhil Kapur

Time Stamps:

01:25 The path to starting LlamaIndex + initial ideas

07:09 LLMs like Chat-GPT vs traditional machine learning

10:00 4 steps of traditional machine learning

10:45 How do large LLMs change the game?

14:11 How does LlamaIndex help LLMs work with unstructured data?

18:08 How do you work with gigabytes of private data?

19:57 Organizing words and paragraphs by topic with embeddings

24:55 The importance of structuring data

26:00 3 key abstractions in LlamaIndex

29:25 Medical use cases for LlamaIndex

31:29 Increasing efficiency in medicine

33:25 An AI medical Research Assistant (Insight)

34:31 Other methods of connecting LLMs to data

36:55 What is langchain?

39:56 What work in the AI and LLM space excites you the most?

42:23 Do you ever feel scared about the developments of AI?

43:45 Llamas and Machine Learning

45:36 What do you think the future of AI in medicine will look like in 10-20 years?

47:24 What advice would you give to grad students, med students, and other early career professionals getting into AI and medicine?