Fall vLLM Meetup Highlights: When AI Builders Unite

Last week, more than 140 AI engineers, researchers, and open-source builders gathered at the vLLM Meetup in Palo Alto to dive into high-performance LLM inference and hands-on AI innovation powered by AMD GPUs.

Hosted at Playground Global, this meetup brought together the brightest minds in the Silicon Valley’s AI community for a rare opportunity to experiment with AMD ROCm™ software, CI/CD pipelines, and interactive Juypyter notebooks, while, exchanging insights and ideas that are shaping the future of AI. With vLLM and AMD leading the way, the meetup showcased how open hardware, software, and collaboration are driving the next generation of AI systems.

Opening Insights:

The evening opened with Ramine Roane’s (Corporate Vice President, AI Product Management, AMD) welcome remarks, setting an energetic tone as he addressed the momentum behind open AI development and the growing community building on the AMD GPUs.  Anush Elangovan (Corporate Vice President of Software Development, AMD) shared updates on the AMD AI Academy and open-source initiatives, reaffirming their commitment to accessible tooling, transparent development, and empowering contributors across the AI ecosystem. 

Spotlight Sessions: Insights from the Experts

The technical program delivered a series of engaging talks:

• Simon Mo, vLLM Project Co-Lead, vLLM: Walked the audience through vLLM’s latest updates, roadmap priorities, and expanding support for AMD ROCm software, emphasizing ease-of-use and performance.
• Lu Fang, Meta: Offered a deep dive into high-performance LLM inference with vLLM, explaining how the architecture scales efficiently while maximizing throughput and memory performance.
• Andy Luo, Senior Director, AMD: Shared practical strategies for optimizing vLLM on AMD GPUs, providing developers with actionable tips for tuning kernels, improving GPU utilization, and unlocking ROCm-accelerated capabilities.
  
  
  

Hands-On Workshop: Building a Two-Agent System

The hands-on workshop brought the evening to life. Eda Zhou (Software Development Engineer, AMD) led a fully booked workshop “Building a Two-Agent System on AMD GPUs (vLLM + MCP),” giving the audience an opportunity to design and deploy multi-agent AI systems in real time.

• 75 developers built live on AMD GPUs; 64 new GPU access requests submitted during the spot.
• Participants explored ROCm workflows, instant Jupyter notebooks, and real-world model-serving environments.
• Developers connected specialized agents through the Model Context Protocol (MCP), deployed open-weight models using vLLM, and created interactive pipelines where agents communicated, coordinated tasks, and produced explainable outputs.

By the end, attendees had a functional two-agent system — a powerful demonstration of how open hardware, open models, and open protocols unlock the next generation of intelligent, action-taking AI systems.

Networking and Community Momentum

The evening transitioned into a lively reception and networking session, where conversations continued well into the evening as engineers, researchers, and open-source contributors exchanged ideas and planned collaborations.

Looking Ahed: What’s Next

The energy in the room proved that the next wave of AI innovation isn’t happening behind closed doors—it’s happening in community meetups, open-source repos, and hands-on workshops like this one.

AMD is excited to carry this momentum forward, strengthening the bridge between developers and the open AI tools that power their ideas. More events, more learning opportunities, and more accessible GPU resources are already on the horizon.

Developer Resources:

• AMD AI Academy - Self-learning courses designed specifically for AI developers.
• ROCm AI Developer Hub - Access SDKs, libraries, documentation, and tools to accelerate AI, HPC, and graphics development.
• AI@AMD X – Stay updated with the latest software releases, AI blogs, tutorials, and news.
• Developer Cloud – Start projects on AMD Instinct™ GPUs with $100 in complimentary credits for 30 days, offering an easy on-ramp for experimentation and benchmarking.
• Developer Central YouTube – Explore hands-on videos, demos, and deep-dive sessions from engineers and community experts.
• Developer Community Discord – Join global developer communities, share feedback, and exchange optimization tips directly with peers and AMD specialists.