Impact AI

• Foundation Model Series: Advancing Precision Medicine in Radiology with Paul Hérent from Raidium

  Radiologists face a growing demand for imaging analysis, yet existing AI tools remain fragmented, each solving only a small part of the workflow. Today, we continue our series on domain-specific foundation models with Paul Hérent, Co-Founder and CEO of Raidium. He joins us to discuss how foundation models could revolutionize radiology by providing a single AI-powered solution for multiple imaging modalities.Paul shares his journey from radiologist to AI entrepreneur, explaining how his background in cognitive science and medical imaging led him to co-found Raidium. He breaks down the challenges of building a foundation model for radiology, from handling massive datasets to addressing bias and regulatory hurdles, and their approach at Raidium. We also explore Raidium’s vision for the future: its plans to refine multimodal AI, expand its applications beyond radiology, and commercialize its technology to improve patient care worldwide. Tune in to learn how foundation models could shape the future of radiology, enhance patient care, and expand global access to medical imaging!Key Points:Paul Hérent’s background in radiology, cognitive science, and founding Raidium.Why existing AI tools in radiology are fragmented and have limited adoption.How Raidium’s foundation model unifies multiple radiology tasks.Raidium’s multimodal AI: handling diverse imaging types in one system.Outlining the vast, diverse data used to train Raidium’s model, including radiology reports.The teams, compute power, and infrastructure behind Raidium’s AI development.Challenges in data curation, regulatory hurdles, and proving clinical value.What makes a good foundation model and the role of self-supervised learning (SSL).Insights into how Raidium benchmarks its model using rigorous medical imaging tests.The role of diverse data, human oversight, and continuous learning in reducing bias.Their current R&D phase and plans for commercialization.Key lessons Paul learned about AI startups, from data needs to product-market fit.The future of foundation models in radiology and beyond.Paul’s advice to AI founders: Build a team with both AI and domain expertise.Raidium’s vision: Improving the lives of patients and global healthcare access.Quotes:“In practice, there is still little AI adoption because every solution solves only a tiny part of what radiologist do. [For radiologists] it's a wider job. We want, as a radiologist, to have one tool to rule all modalities.” — Paul Hérent“Data is key. If you have good data, not only to build a data set, but proprietary data, challenging data, rare data in a specific domain. It's very valuable because the architecture is not particularly innovative.” — Paul Hérent“Build a team with people you trust. Entrepreneurship is not trivial. Be complementary.” — Paul Hérent“The dream of Raidium is to build something that has a huge impact on a patient's life.” — Paul Hérent“If we go beyond the rich countries, many, many people have no access to radiology. Two-thirds of countries don’t have access to radiologists. It's a big need. If we can contribute with our approach to more accessible health, we will be very happy.” — Paul HérentLinks:Paul Hérent on LinkedInPaul Hérent on Google ScholarRaidiumResources for Computer Vision Teams:LinkedIn – Connect with Heather.Computer Vision Insights Newsletter – A biweekly newsletter to help bring the latest machine learning and computer vision research to applications in people and planetary health.Computer Vision Strategy Session – Not sure how to advance your computer vision project? Get unstuck with a clear set of next steps. Schedule a 1 hour strategy session now to advance your project.

  --------  

  22:36
• Foundation Model Series: Advancing Endoscopy with Matt Schwartz from Virgo

  What if a routine endoscopy could do more than just detect disease by actually predicting treatment outcomes and revolutionizing precision medicine? In this episode of Impact AI, Matt Schwartz, CEO and Co-Founder of endoscopy video management and AI analysis platform Virgo, discusses how AI and machine learning are transforming endoscopy.Tuning in, you’ll learn how Virgo’s foundation model, EndoDINO, trained on the largest endoscopic video dataset in the world, is unlocking new possibilities in gastroenterology. Matt also shares how automated video capture, AI-powered diagnostics, and predictive analytics are reshaping patient care, with a particular focus on improving treatment for inflammatory bowel disease (IBD). Join us to discover how domain-specific foundation models are redefining healthcare and what this means for the future of precision medicine!Key Points:An introduction to Matt Schwartz and Virgo’s mission.The importance of video documentation in endoscopy and its impact on healthcare.Machine learning’s role in automating endoscopic video capture and clinical trial recruitment.Building the EndoDINO foundation model to unlock endoscopy data for precision medicine.Data collection: the process of gathering 130,000+ procedure videos for model training.Foundation model development using self-supervised learning and DINOv2.Model development challenges, from hyper-parameter tuning to domain-specific adjustments.Applying EndoDINO to predict inflammatory bowel disease (IBD) treatment responses.Commercializing EndoDINO through licensing to health systems and pharma companies.The future of foundation models in endoscopy: expanding applications beyond GI diseases.Advice for AI startup founders to prioritize data capture as a foundation for AI success.Insight into Virgo’s vision to transform IBD treatment and preventative care.Quotes:“There's a massive amount of endoscopic video data being generated across a wide range of endoscopic procedures, and nobody was capturing that data – [Virgo] realized early on that endoscopy data could hold the key to unlocking all sorts of opportunities in precision medicine.” — Matt Schwartz“With the foundation model paradigm, you can compress a lot of heavy compute needs into a single model and then build different applications on top of the foundation. This is going to have a positive impact on the clinical deployment of foundation models.” — Matt Schwartz“Our foundation model can turn something like a routine colonoscopy into a precision medicine screening tool for IBD patients.” — Matt Schwartz“There are a lot of untapped data resources in healthcare. If a founder can build a first product that is the data capture engine, it will set them up for a ton of future success when it comes to AI development.” — Matt SchwartzLinks:VirgoMatt Schwartz on LinkedInMatt Schwartz on XEndoMLIntroducing EndoDINO: A Breakthrough in Endoscopic AIResources for Computer Vision Teams:LinkedIn – Connect with Heather.Computer Vision Insights Newsletter – A biweekly newsletter to help bring the latest machine learning and computer vision research to applications in people and planetary health.Computer Vision Strategy Session – Not sure how to advance your computer vision project? Get unstuck with a clear set of next steps. Schedule a 1 hour strategy session now to advance your project.

  --------  

  21:04
• Foundation Model Series: Transforming Biology with Zelda Mariet from Bioptimus

  Zelda Mariet, Co-Founder and Principal Research Scientist at Bioptimus, joins me to continue our series of conversations on the vast possibilities and diverse applications of foundation models. Today’s discussion focuses on how foundation models are transforming biology. Zelda shares insights into Bioptimus’ work and why it’s so critical in this field. She breaks down the three core components involved in building these models and explains what sets their histopathology model apart from the many others being published today. They also explore the methodology for properly benchmarking the quality and performance of foundation models, Bioptimus’ strategy for commercializing its technology, and much more. To learn more about Bioptimus, their plans beyond pathology, and the impact they hope to make in the next three to five years, tune in now.Key Points:Who is Zelda Mariet and what led her to create Bioptimus. What Bioptimus does and why it’s so important.Why their first model announced was for pathology.Zelda breaks down three core components that go into building a foundation model.How their histopathology foundation model is different from the number of other models published at this point.Their methodology behind properly benchmarking how well their foundation model performs.Different challenges they’ve encountered on their foundation model journey.How they plan to commercialize their technology at Bioptimus. Thoughts on whether open source is part of their long-term strategy for the model, and why.  Developing a product roadmap for a foundation model.She shares some information regarding their next step, beyond pathology, at Bioptimus.The importance of understanding what kind of structure you want to capture in your data.Where she sees the impact of Bioptimus in the next three to five years. Quotes:“Working on biological data became a little bit of a fascination of mine because I was so instinctively annoyed at how hard it was to do.” — Zelda Mariet“Bioptimus is building foundation models for biology. Foundation models are essentially machine learning models that take an extremely long time to train [and] are trained over an incredible amount of data.” — Zelda Mariet“There are two things that are well-known about foundation models, they’re hungry in terms of data and they’re hungry in terms of compute.” — Zelda Mariet“On the philosophical side, science is something that progresses as a community, and as much as we have, what I would say is a frankly amazing team at Bioptimus, we don’t have a monopoly on people who understand the problems we’re trying to solve. And having our model be accessible is one way to gain access into the broader community to get insight and to help people who want to use our models, get insight into maybe where we’re not doing as well that we need to improve.” — Zelda MarietLinks:Zelda Mariet on LinkedInZelda MarietBioptimusResources for Computer Vision Teams:LinkedIn – Connect with Heather.Computer Vision Insights Newsletter – A biweekly newsletter to help bring the latest machine learning and computer vision research to applications in people and planetary health.Computer Vision Strategy Session – Not sure how to advance your computer vision project? Get unstuck with a clear set of next steps. Schedule a 1 hour strategy session now to advance your project.

  --------  

  21:26
• Foundation Model Series: Democratizing Time Series Data Analysis with Max Mergenthaler Canseco from Nixtla

  What if the hidden patterns of time series data could be unlocked to predict the future with remarkable accuracy? In this episode of Impact AI, I sit down with Max Mergenthaler Canseco to discuss democratizing time series data analysis through the development of foundation models. Max is the CEO and co-founder of Nixtla, a company specializing in time series research and deployment, aiming to democratize access to advanced predictive insights across various industries.In our conversation, we explore the significance of time series data in real-world applications, the evolution of time series forecasting, and the shift away from traditional econometric models to the development of TimeGPT. Learn about the challenges faced in building foundation models for time series and a time series model’s practical applications across industries. Discover the future of time series models, the integration of multimodal data, scaling challenges, and the potential for greater adoption in both small businesses and large enterprises. Max also shares Nixtla’s vision for becoming the go-to solution for time series analysis and offers advice to leaders of AI-powered startups.Key Points:Max's background in philosophy, his transition to machine learning, and his path to Nixtla.Why time series data is the “DNA of the world” and its role in businesses and institutions.Nixtla's advanced forecasting algorithms, the benefits, and their application to industry.Historical overview of time series forecasting and the development of modern approaches.Learn about the advantages of foundation models for scalability, speed, and ease of use.Uncover the range of datasets used to train Nixtla's foundation models and their sources.Similarities and differences between training TimeGPT and large language models (LLMs).Hear about the main challenges of building time series foundation models for forecasting. How Nixtla ensures the quality of its models and the limitations of conventional benchmarks.Explore the gap between benchmark performance and effectiveness in the real world.He shares the current and upcoming plans for Nixtla and its TimeGPT foundation model. He shares his predictions for the future of time series foundation models.Advice for leaders of AI-powered startups and what impact he aims to make with Nixtla.Quotes:“Time series are in one aspect, the DNA of the world.” — Max Mergenthaler Canseco“Time is an essential component to understand a change of course, but also to understand our reality. So, time series is maybe a somewhat technical term for a very familiar aspect of our reality.” — Max Mergenthaler Canseco“Given that we are all training on massive amounts of data and some of us are not disclosing which datasets we’re using, it’s always a problem for academics to try to benchmark foundation models because there might be leakage.” — Max Mergenthaler Canseco“That’s an interesting aspect of foundation models in time series, that benchmarking is not as straightforward as one might think.” — Max Mergenthaler Canseco“I think right now in our field probably benchmarks are not necessarily indicative of how well a model is going to perform in real-world data.” — Max Mergenthaler Canseco“I think that we’re also going to see some of those intuitions that come from the LLM field translated into the time series field soon.” — Max Mergenthaler CansecoLinks:Max Mergenthaler Canseco on LinkedInNixtlaNixtla on XNixtla on LinkedInNixtla on GitHubResources for Computer Vision Teams:LinkedIn – Connect with Heather.Computer Vision Insights Newsletter – A biweekly newsletter to help bring the latest machine learning and computer vision research to applications in people and planetary health.Computer Vision Strategy Session – Not sure how to advance your computer vision project? Get unstuck with a clear set of next steps. Schedule a 1 hour strategy session now to advance your project.

  --------  

  27:11
• Foundation Model Series: Harnessing Multimodal Data to Advance Immunotherapies with Ron Alfa from Noetik

  In this episode, I'm joined by Ron Alfa, Co-Founder and CEO of Noetik, to discuss the groundbreaking role of foundation models in advancing cancer immunotherapy. Together, we explore why these models are essential to his work, what it takes to build a model that understands biology, and how Noetik is creating and sourcing their datasets. Ron also shares insights on scaling and training these models, the challenges his team has faced, and how effective analysis helps determine a model’s quality. To learn more about Noetik’s innovative achievements, Ron’s advice for leaders in AI-powered startups, and much more, be sure to tune in!Key Points:Ron shares his background and how his journey led to Noetik.Why a foundation model is important in their work.What goes into building a foundation model that understands biology.Building the dataset: where does the data come from?The types of data they generate from the samples they use in their models.He further explains the components necessary to build a foundation model.The scale and what it takes to train these models. Ron sheds light on the challenges they’ve encountered in building their foundation model.How to determine if your foundation model is good. Utilizing analysis to help identify ways to improve your model. The current purpose for their foundation model and how they plan to use it in the future.Key insights gained from developing foundation models and how these can be adapted to other types of data.His advice to other leaders of AI-powered startups.Ron digs deeper into their goal to impact patient care by developing new therapeutics.Quotes:“Our thesis for Noetik is that one of the biggest problems we can impact if we want to make and bring new drugs to patients is predicting clinical success; so-called translation — that's where we focus Noetik, how can we train foundation models of biology so that we can better translate therapeutics from early discovery and preclinical models to patients.” — Ron Alfa“We think the most important thing for any application of machine learning is the data.” — Ron Alfa“The goal here is to train models that can do what humans cannot do, that can understand biology that we haven't discovered yet.” — Ron Alfa“The big aim of Noetik is to develop these [foundational] models for therapeutics discovery.” — Ron AlfaLinks:Ron Alfa on LinkedInRon Alfa on XNoetikNoetik Octo Virtual Cell (OTCO)Resources for Computer Vision Teams:LinkedIn – Connect with Heather.Computer Vision Insights Newsletter – A biweekly newsletter to help bring the latest machine learning and computer vision research to applications in people and planetary health.Computer Vision Strategy Session – Not sure how to advance your computer vision project? Get unstuck with a clear set of next steps. Schedule a 1 hour strategy session now to advance your project.

  --------  

  33:53

Plus de podcasts Technologies

Podcasts tendance de Technologies

À propos de Impact AI