Join us for the Red Hat AI Day of Learning, a virtual event designed for developers, engineers, and technical practitioners who want to deepen their expertise in AI inference, model optimization, model customization, agentic AI, and more.
At the Red Hat AI Day of Learning, you’re in control of your learning path. With 12 breakout sessions across four different tracks, you can mix and match sessions to focus on the areas most relevant to you:
- Learn how to optimize inference performance with vLLM, LLM Compressor, and Speculators.
- Explore ways to connect models to enterprise data through end-to-end model customization, synthetic data generation, training workflows, and more.
- Dive into agentic AI development with MCP, LlamaStack, and open source agent frameworks.
- See how to scale AI across the hybrid cloud with distributed inference and Red Hat AI on OpenShift AI.
P.S. Don’t miss our “What’s New and What’s Next in Red Hat AI” product update on October 14th, featuring Red Hat AI’s leadership team.
Track 1: Increase Efficiency with Fast, Flexible, and Efficient Inferencing
10:00 am – Intro to vLLM and how to get started (30 mins)
Description: A practical introduction to running and deploying models with vLLM, the de-facto open source inference engine for LLMs.
Speaker:
- Michael Goin, Principal Software Engineer, Red Hat, and vLLM Core Committer
10:30 am – Model optimization with LLM Compressor (30 mins)
Description: Learn the importance of compressing models and how to accurately compress LLMs for fast and efficient inference with vLLM, using the open source LLM Compressor library.
Speakers:
- Dipika Sikka, ML Technical Lead, Red Hat
- Kyle Sayers, Sr. Machine Learning Engineer, Red hat
11:00 am – Lossless LLM inference acceleration with Speculators (30 mins)
Description: Meet the Speculators library! Create and fine-tune speculative decoding for your use cases, test and iterate to cut latency, and deploy in vLLM for up to 3X faster inference without loss in quality.
Speakers:
- Mark Kurtz, Member of Technical Staff, Red Hat
- Megan Flynn, Senior Machine Learning Engineer, Red Hat
Track 2: Simplified and Consistent Experience for Connecting Models to Data
10:00 am – End-to-end model customization (30 mins)
Description: Learn what model customization is and when it's the right approach for your use case. This session covers the spectrum of customization techniques—from prompt engineering to fine-tuning of your models—and explores different scenarios where each approach is most effective. We'll look at real-world examples where customization delivers significant value, such as adding domain-specific knowledge through knowledge tuning or improving efficiency by converting long prompts into fine-tuned behaviors. The session includes a brief demonstration of synthetic data generation and training as one pathway to model customization. This foundational overview prepares you for deeper dives into specific techniques in follow-up sessions.
Speakers:
- Kai Xu, Engineering Manager and Senior Principal Research Scientist, Red Hat
10:30 am – Synthetic data generation and data processing (30 mins)
Description: Deep dive into synthetic data generation. Explore how to generate and process synthetic data with tools like Docling, apply subset selection techniques, and leverage SDG Hub to streamline model customization workflows.
Speakers:
- Shiv Sudalairaj, Principal Research Scientist, Red Hat
- Hao Wang, Research Scientist, Red Hat
11:00 am – Training models: Continual learning of LLMs with Training Hub (30 mins)
Description: Discover approaches for training large language models, explore continual learning techniques, and see how Training Hub supports scalable, iterative model development.
Speaker:
- Mustafa Eyceoz, Principal Research Scientist, Red Hat
Track 3: Accelerate Agentic AI Delivery and Stay at the Forefront of Innovation
10:00 am – Build open source agentic AI solutions (30 mins)
Description: Learn how to design and implement agentic AI systems with open source tools, focusing on flexibility, transparency, and community-driven innovation.
Speakers:
- Younes Ben Brahim, Principal Product Marketing Manager, Red Hat
- Adel Zaalouk, Principal Product Manger, Red Hat
10:30 am – Intro to Model Context Protocol (MCP) (30 mins)
Description: Dive into the Model Context Protocol, a standard for connecting models with external tools and data sources, enabling more capable and context-aware AI agents.
Speakers:
- Cedric Clyburn, Sr. Developer Advocate, Red Hat
- Peter Double, AI Principal Product Manager, Red Hat
11:00 am – Intro to Llama Stack (30 mins)
Description: Explore Llama Stack, an open framework for building, scaling, and customizing agent workflows. Learn how it makes experimentation and deployment of agentic AI in production easier.
Speakers:
- Phillip Hayes, Senior AI Platform Architect, AI Customer Adoption and Innovation, Red Hat
- Roberto Carratalá, Principal AI Platform Architect, Red Hat
Track 4: Flexibility and Consistency When Scaling AI Across the Hybrid Cloud
10:00 am – Intro to distributed inference (30 mins)
Description: Get an overview of distributed inference, including how to scale model serving across multiple GPUs, balance workloads efficiently, and achieve higher throughput and lower latency for large-scale deployments.
Speaker:
- Christopher Nuland, Technical Marketing Manager, Red Hat
10:30 am – Distributed inference with llm-d’s “well-lit paths” (30 mins)
Description: Introduction to llm-d, a Kubernetes-native high-performance distributed LLM inference framework. Dig into llm-d's “well-lit paths” approach to distributed inference, featuring intelligent inference scheduling, prefill/decode disaggregation, and efficient scaling for very large Mixture of Expert (MoE) models like DeepSeek-R1.
Speaker:
- Robert Shaw, Member of Technical Staff, Red Hat
11:00 am – Solving the scaling challenge: Three proven strategies for your AI infrastructure
Description: Explore how platform teams can empower data practitioners and AI engineers to work efficiently while ensuring scalability, observability, and automation. Get familiar with three customer scenarios—maximizing GPU utilization, gaining control over open source models, and scaling inference, all three complete with demos, guidance, and GitHub links to help you apply these approaches on Kubernetes-based infrastructure.
Speakers:
- Will McGrath, Sr. Principal Product Marketing Manager, Red Hat
- James Harmison, Senior Principal Technical Marketing Manager, Red Hat
- Phillip Hayes, Senior AI Platform Architect, AI Customer Adoption and Innovation, Red Hat
