NVIDIA and the Healthcare & Life Sciences Startups Ecosystem
A virtual event designed for healthcare developers and startups, this summit offers a full day of technical talks to reach developers and technical leaders in the EMEA region. Get best practices and insights for applications, from biopharma to medical imaging.
Speakers
Jaap Zuiderveld
Vice President, EMEA
NVIDIA
Renee Yao
Global Healthcare AI Startup Lead
NVIDIA
Chris Dallago
Senior Solutions Architect
NVIDIA
Lindsay Edwards
CTO and President of Platform
RelationRX
Andres Diaz-Pinto
Technical Marketing Engineer
NVIDIA
Tal Einat
Software Engineering Lead
Rhino Health
Jorrit Glastra
Chief Technology Officer
Quantib
Mattia Lupetti
Senior Research Scientist
ImFusion
Harry Clifford
Product Manager for Clara Parabricks
NVIDIA
Craig Rhodes
Healthcare & Life Sciences Industry Leader
NVIDIA
Bruno Trentini
Developer Relations Manager Healthcare & Life Sciences Startups EMEA
NVIDIA
Sessions
- Generic Sessions
- Natural Language Processing (NLP) & Large Language Models (LLM)
- Medical Imaging Advancements
- Genomics
NVIDIA and the Healthcare & Life Sciences Startups Ecosystem
Learn why now is an unprecedented time in computational biology and drug discovery, and the areas in which NVIDIA and the ecosystem are innovating to bring forth these new capabilities. Join Jaap Zuderveild, VP EMEA, and Renee Yao, Global Healthcare AI Startups Lead, in this session to get an overview about the tools, platforms and services NVIDIA created to support developers in the Healthcare and Life Sciences domain, from Drug Discovery to Medical Imaging thinking about engineers, scientists and technical leaders. We'll highlight examples of recent breakthroughs across life sciences, explore the methods and techniques used to enable such advancements, and bring awareness to relevant libraries, frameworks, models, and applications for developers to be aware of.
Hassle-Free Compute Scalability For Startups And Enterprises
Startups often face many challenges in scaling their compute requirements given high costs associated with training models. A moderately-sized model can cost millions of dollars to train in the most popular cloud instances, causing business and technical turmoil to Startups, especially those who lack dedicated funding to scale compute. In this session we will chat about alternatives available to Startups in the EMEA region, including hybrid approaches and superPOD rental models, which can reach hundreds of Petaflops of AI performance at a fraction of a cost. We’ll also provide advice to help engineering leads navigate both common and overlooked challenges based on our industry experiences.
Protein Language Models For Protein Science, Design And Engineering
Biology has recently seen incredible advances supported by methods from Deep Learning. In particular, protein science and engineering has benefitted both from advances in downstream deep learning models, predicting aspects of protein structure and function, and foundational models often based on the Transformer architecture, providing new ways to encode protein information in computational form. The staggering advances in accurate Deep Learning methods for protein science have led to a sprouting of new ways in which Deep Learning can be used to predict and design proteins. This talk will examine the principles that led to foundational models for protein science, the validation of these models through predictions of function and structure, and venues for their application in protein engineering and design accompanied by practical examples.
Learning Gene Cis-Regulation Using LLM And Cambridge-1
Since 2016, scientists have been using neural sequence models to model and understand DNA (and latterly protein) sequences. Initially using 1D-CNNs, then LSTMs, scientists (including those at Relation) now use Transformer-based architectures to predict how changes in DNA sequence affect the biology of the DNA itself (for example, whether key molecules called transcription factors will bind or not). Ultimately, these models can predict how changes in DNA sequence might affect gene expression. In this talk I will outline the science underlying these efforts in detail, before describing Rosalind - Relation’s large language model for DNA - and the role of Cambridge-1 in pushing the scientific bounds in this exciting and important area of biology.
BioNeMo and NVIDIA Life Sciences NLP
Transformers have emerged as perhaps the most fundamental AI technologies of our time, initially in their application to natural language processing (NLP) and the remarkable capabilities that emerge when very large models are trained on huge, diverse datasets. These capabilities reveal that, in being trained on seemingly mundane tasks – such as predicting the identity of a masked token or the next token in a sequence – these models capture not only the structure and meaning of language, but learn to wield it for function and purpose. When applied to the language of biology itself, represented in the sequences of proteins that fold to do useful work, or the expression and transcription of our DNA, transformers have similarly demonstrated remarkable capacity to help predict the structure, properties, and function of biomolecules. NVIDIA is building state-of-the-art models and a framework to support this revolution, called BioNeMo, which extends the leadership of NVIDIA’s NLP technologies into the language of biology and chemistry. In this talk, we will introduce and review the fundamental technologies used to scale all NVIDIA life sciences NLP with a special focus on BioNeMo, supported architectures, and a window into the roadmap of these technologies.
MONAI Label: A Framework For Interactive Image Annotation Through AI
MONAI Label is an open-source image labeling and learning tool that allows researchers to create novel AI models and collaborate with a clinical team. In this talk, you’ll get a short introduction to MONAI Label and then dive deep into utilizing OHIF and 3D Slicer as your viewers and use an existing application to start with AI-assisted Annotation. Next, you’ll learn more about the continuous learning loop integrated into MONAI Label and how active learning plays an integral role in data selection. You’ll also learn how to create your own custom MONAI Label sample application so you can understand how to integrate your workflow into MONAI Label.
Privacy-Preserving AI in Healthcare
Federated Learning is a promising approach to unlocking collaborative, large-scale medical research and development, without compromising patient data privacy. However, there are still challenges slowing its widespread adoption. We will discuss these challenges, potential solutions, and present the Rhino Health Platform, an end-to-end privacy-preserving solution for distributed computing in healthcare. We will provide a practical guide for every part of distributed digital health projects. Specifically, we will demonstrate our integration of NVIDIA FLARE into the platform as a flexible framework for training AI models via Federated Learning.
Accelerating Deep Learning Model Development For Radiology With MONAI
It is clear that in radiology image interpretation AI and in particular deep learning can add value to human observers. The development of deep learning models to support radiology tasks including detection, segmentation and quantification has risen strongly over recent years with certified applications making their way into clinical practice. The particular difficulty of processing 3D images has led to a steady flow of new ideas from computer vision and general deep learning being applied to radiology problems. For device developers to maximize the impact of these innovations, it is key to rapidly assess and, if beneficial, adopt these improvements. Leveraging an industry best practice code base, like MONAI core, that follows the field can enable this. Better models in turn tend to expose areas where training labels are incorrect or inconsistent. Continuously enhancing the level of annotations is a second key component of high quality model development. Rapid iteration between models and expert humans as enabled by MONAI label helps unlock this approach effectively.
The Challenges Of Medical Imaging R&D In The Era Of Deep Learning
As an R&D service provider in the field of medical image analysis, ImFusion has long embraced the AI revolution and made Deep Learning a fundamental building block of our technology stack. Its tight integration within our GPU-accelerated SDK allows us to effectively tailor our software solutions to a variety of clinical applications, including performance-critical interventional scenarios. In this talk we'll discuss the challenges and the lessons learned in the development and deployment of DL-based medical image processing workflows from the perspective of an R&D consulting startup.
Deep Learning and Acceleration In Genomics On GPUs
For precision medicine to become routine, genome sequencing needs to be delivered at high speed, low cost, and at scales that drive new understandings in human biology and medicine. GPU-based parallel processing and sequencing data, in addition to accelerated deep learning in base calling and variant calling, is now addressing the myriad bottlenecks that occur across the computational workflow. With performance of up to 80x acceleration for state-of-the-art bioinformatics tools and complete end-to-end workflows in under 25 minutes, larger sequencing projects are becoming less expensive, easier to manage, and generating more useful insights than ever before. NVIDIA is delivering this accelerated compute framework for end-to-end genome sequencing analysis. This talk will provide an overview of NVIDIA's Clara Parabricks software for accelerated analysis and deep learning in genomics, including several use cases from critical care rapid sequencing to large-scale population studies.
Featured Inception Start-Ups
Featured Resources
MONAI for Medical Imaging
Born of a network of healthcare AI thought leaders, MONAI addresses the full stack for medical AI, including labeling, training, inference and deployment.
BioNeMo Early Access Program
BioNeMo is an AI-powered drug discovery cloud service and framework built on NVIDIA NeMo Megatron for training and deploying large biomolecular transformer AI models at supercomputing scale.
Healthcare Developer Resource
A hub of technical resources, news, and SDKs for developers and data scientists working in the Healthcare & Life Sciences industry.
Connect with Top Investors
NVIDIA gives Premier members and other select startups unique opportunities to interact with our worldwide ecosystem of investors. Learn about the NVIDIA Inception VC Alliance, an initiative between NVIDIA Inception and top-tier venture capital firms.
Stay up to date with the latest healthcare news from NVIDIA