The Hard Part of Machine Learning with Lynn Langit
Jun 26, 2024
auto_awesome
Expert Lynn Langit discusses the challenges of using machine learning to improve patient outcomes at Mayo Clinic. She highlights the complexities of multi-modal data analytics, testing machine learning models, and the importance of focusing on fundamentals before advanced goals. The conversation explores practical applications in bioinformatics, statistical tests for machine learning, and the impact of new technologies like Gemini in Google Cloud on productivity and collaboration.
Challenges of multi-modal data analytics include integrating diverse patient data for comprehensive treatment evaluation.
Practical machine learning success relies on mastering fundamentals before advancing to complex tasks.
Deep dives
Lynn Langit's Versatility Across Cloud Platforms
Lynn Langit, a noted cloud architect working with GCP, AWS, and Azure, is known for her technical expertise and customer-centric approach. Her recognition as a Google developer expert in cloud and AI, an AWS community hero for data, and a Microsoft regional director underscores her agnostic stance, prioritizing serving customers across diverse platforms for optimal solutions.
Venturing into the Cloud with Mayo Clinic and the Broad Institute
Langit's involvement with the cloud spans major research institutions like the Mayo Clinic and the Broad Institute at MIT and Harvard. Focusing on bioinformatic research, her work contributes to groundbreaking advancements in applied medicine, especially evident in the Mayo Clinic's 10-year cloud migration journey and breakthroughs in heart transplants and medical achievements.
Challenges in Implementing Machine Learning Models at Mayo Clinic
Working closely with Mayo Clinic's bioinformaticians and biostatistics professionals, Langit delves into the complexities of transitioning from traditional machine learning methods to generative AI initiatives. The application of large language models (LLMs) in medicine poses challenges due to statistical differences and low confidence responses, requiring careful thresholds and adaptation processes.
The Significance of Multimodal Data in Healthcare Analysis
Exploring the advantages of multimodal data analysis, Langit highlights the importance of integrating various data sources like patient notes, proteins expression, and medical images. By combining traditional ML models with LLMs and embracing spatial multiomics, healthcare professionals aim to enhance clustering and treatment outcomes, heralding a new era of data-driven patient care.
What are the hard parts of machine learning? Richard chats with Lynn Langit about her work helping the Mayo Clinic improve patient outcomes using machine learning to understand patient data better. Lynn talks about the challenges of multi-modal data analytics - taking all the different data collected from a patient, like an X-ray or video, along with treatment notes, to create an overall picture of treatment and outcome. Then multiply that by thousands of patients, making a complicated data problem with huge challenges in testing and validation. How do you know that the machine learning model is correct? The key to practical machine learning is in the fundamentals - working on each step before you jump to the more complex goals!