Frank Kraemer is an IBM Systems Architect for large-scale IT solutions. He has a very good understanding of Data Management for Autonomous Driving (AD) and AI in the Automotive Industry.
Frank Kraemer is responsible for helping drive global automotive industry enterprise data strategy and focused on establishing areas of application for the Red Hat OpenShift, full-stack and software-defined computing platform used in areas of deep learning, data science, and artificial intelligence.
Presentation
AI for Autonomous Driving (AD) Needs a Solid Data Platform
From ChatGPT to generative world models for autonomous driving (AD), the integration of large-scale AI models into automotive software development is becoming increasingly prevalent. These models range from vision AI models to end-to-end AI models for autonomous driving.
AI and scientific-computing applications are strong examples of distributed-computing problems. The workloads are too large and the computations too intensive to run on a single machine. Computations are therefore broken down into parallel tasks distributed across thousands of compute engines, such as CPUs and GPUs. To achieve scalable performance, systems divide workloads such as training data, model parameters, or both across multiple nodes. These nodes must frequently exchange information, including gradients from newly processed model computations during backpropagation in model training, requiring efficient collective communications such as all-reduce, broadcast, gather and scatter operations.
Unlike traditional applications that work with structured data, performance-intensive AI and analytics workloads operate on unstructured data such as sensor data, audio, images, videos and other objects. Data for AI model training must be high-performance, flexible and scalable. Data enables AI models to identify and extract meaningful features from input data, while the quality and depth of training data significantly influence the success of AI models. Training data provides the examples and relevant information from which AI models learn.
Audience will learn:
#1: Data for AI model training must be high-performance, flexible and scalable.
#2: The quality and depth of training data significantly impact the success of AI models.
#3: Data helps AI models identify and extract meaningful features from input data.
