Machine Learning on GPUs is a one-day, intensive course that teaches attendees how to accelerate data science workflows using GPUs. It covers GPU and CUDA fundamentals, GPU-accelerated data processing, tabular machine learning with cuML, and neural network acceleration techniques. These topics give participants the skills to use GPUs effectively at each stage of the data science process. Attendees get hands-on experience with tools such as Numba, cuDF, cuML, and PyTorch.
The opening module grounds attendees in general-purpose GPU compute before digging into the CUDA compute model itself. It also covers the driver and compute-version compatibility issues that trip up real deployments. Data processing labs move data onto the GPU with Numba and cuDF, then scale that work across multiple GPUs with Dask-RAPIDS. The tabular ML module has attendees train, evaluate, and persist models with cuML, then scale training with Dask-RAPIDS as dataset size grows.
The closing neural network module covers quantization and model compilation for faster inference. It also covers attention acceleration techniques for transformer-style architectures, and how to build data loaders that keep a GPU fed instead of sitting idle. The course runs in a single day, so each module moves quickly from concept to lab. Developers, analysts, data scientists, and MLOps staff alike leave with a practical sense of where GPU acceleration pays off in a real pipeline.
Who Should Attend
Developers, Analysts & Data Scientists, ML & Data Engineers, MLOps & DevOps staff, Technical Managers
What Attendees Will Learn
Upon completing Machine Learning on GPUs, participants will be able to:
- Explain the role of GPUs and CUDA in machine learning and scientific computing
- Process data on the GPU using Numba, cuDF, and Dask-RAPIDS
- Train, evaluate, and persist tabular ML models with cuML
- Accelerate neural networks using quantization and model compilation
- Build effective data loaders and apply attention acceleration techniques
- Make the most of GPU hardware when training neural networks
Prerequisites
Participants should have programming skills (preferably in Python), an understanding of mathematics and statistics, a familiarity with machine learning and data science.