[SF] AWS x NVIDIA Workshop for Model Customization

Join us for a hands-on workshop on model customization with AWS and NVIDIA. As companies move from AI prototypes to production, off-the-shelf models may not meet the accuracy, latency, and cost requirements. Whether it's an agent that drops tools, a chatbot that hallucinates, the model needs to learn your data, your rules, your domain. Fine-tuning a smaller open-weight model like Nemotron 3 on targeted tasks can match or exceed much larger proprietary models, delivering real cost savings while keeping your data inside your own secure infrastructure. This workshop gives you the hands-on experience to make that happen, and you'll leave with a customized model and a follow-up plan in hand. What You'll Learn: Apply a customization framework to decide when to prompt-engineer vs. fine-tune Run a supervised fine-tuning (SFT) job on Nemotron 3 using Amazon Sagemaker AI Learn how model customization can improve agentic systems Understand when and how to use different types of fine-tuning techniques Evaluate your customized model against your baseline using custom metrics and benchmark evaluations Deploy your fine-tuned model to a SageMaker real-time endpoint Featured Speakers:

Guy Farber, Global Head of GTM at NVIDIA

Balaji Chandrasekaran, VP of Product at Protopia AI

Curtice Lockhart, Senior AI Solutions Architect at NVIDIA

Ganesh Hedge, Sr. WW Specialist, Frontier Models at AWS

Deepak Rupakula, Principal WorldWide Specialist, Fronteir Models at AWS Workshop Leads: Chakra Nagarajan, Sr. Mgr Specialist SA at AWS Amin Dashti, Sr. GenAI Data Scientist at AWS Sandeep Raveesh-Babu. Sr. WW GenAI Specialist at AWS

Who Should Attend ML engineers and data scientists building agentic or production AI workflows ML/AI leaders evaluating or deploying custom models in production Platform/infrastructure engineers responsible for training and inference at scale Prerequisites To get the most out of this session:

1. Bring a laptop (instructions provided upon registration)
2. Bring your hardest use case - we'll map it to a customization path during the workshop