AMD AI DevDay 2025

Driss Guessous

Machine Learning Engineer, Meta Platforms

Driss is a machine learning engineer working on core development of PyTorch. He received his master's degree in computer science from the University of Illinois at Urbana-Champaign, and a dual degree in physics and applied mathematics from The Ohio State University. Driss also won back-to-back NCAA Division 1 National Championships with the OSU Men's Volleyball team.

Simran Arora

Academic Partner of Togerther AI and PhD Student at Stanford

Simran is a PhD student in Computer Science at Stanford, and is advised by Chris Ré. She is part of the Hazy Research Lab and supported by the Stanford Graduate Fellowship as the Sequoia Capital Fellow. She is also an advisor to Cartesia AI and an academic partner of Together AI.

Her research is in AI systems and focuses on expanding the Pareto frontier between quality and efficiency, to unlock new AI capabilities, by considering AI algorithms, hardware, and applications in lockstep.

Chenyu Zhao

Co-founder and engineer, Fireworks AI​

Chenyu Zhao is a co-founder and engineer at Fireworks AI since 2022. Previously, Chenyu Zhao served as a senior staff software engineer at Google from 2013 to 2022, leading a team of approximately 50 on Google Cloud Vertex AI, with a focus on advanced algorithms like AutoML and MLOps. Earlier experience includes roles as a student instructor and researcher at UC Berkeley, a software engineering intern at NVIDIA and the Federal Reserve Board, and research positions at the University of Maryland, College Park, and the National Institute of Standards and Technology. Chenyu Zhao obtained a B.S. in Electrical Engineering and Computer Science with a focus on Engineering Math and Statistics from the University of California, Berkeley from 2010 to 2013.

Erwan Menard

Senior Vice President of Product Management, Crusoe

Erwan Menard is the Senior Vice President of Product Management at Crusoe, where he leads the innovation and scaling of Crusoe Cloud to advance sustainable AI infrastructure solutions. Before joining Crusoe, Menard was the Director of Product Management for Cloud AI at Google Cloud. In this role, he led the outbound product management for the Vertex AI portfolio, which included the Vertex AI Platform, Search, and Agentspace application products. His team was pivotal in enabling customers to build advanced AI and GenAI-powered experiences and in shaping Google Cloud's AI partnerships strategy.

Prior to his time at Google, Menard was the CEO of Elastifile, a software-defined storage company that was acquired by Google. The technology from Elastifile is now integral to the Google Cloud Filestore product line. He also played a key role in building Google Cloud Solution Engineering, developing programs for migration, application modernization, and data center transformation. Earlier in his career, Menard held executive positions at several global startups and had leadership roles at Alcatel and Hewlett-Packard, where he was the VP & GM of HP's worldwide telecom business unit.

Daniel Han

CEO, Unsloth AI

Daniel Han is the CEO of Unsloth AI, an open-source AI startup on a mission to make AI accessible and accurate for everyone. Unsloth has grown rapidly, earning over 38,000 GitHub stars and surpassing 10 million monthly downloads on Hugging Face. The team has contributed critical bug fixes to major models, including LLaMA, Phi, Gemma, and Mistral. Before founding Unsloth, Daniel worked at NVIDIA, achieving a 2000x speedup in t-SNE visualization.

Simon Mo

vLLM Project Co-Lead, vLLM

Simon Mo is a PhD student at Berkeley Sky Computing Lab. He is a co-lead of vLLM project, which focuses on building an easy to use and highly performance LLM inference on data center hardwares. vLLM is widely adopted, community maintained, and fully open source.

Tris Warkentin

Director of Product Management
Google DeepMind

Tris Warkentin leads Product Management efforts for Google DeepMind’s next-generation AI research, working to achieve Artificial General Intelligence (AGI) and open models (Gemma). Prior to DeepMind, Tris led the Product Management team for Google Brain, the fundamental AI research lab within Google Research. In Brain, he led critical launches in many areas, including Bard, PaLM, LaMDA, Imagen, Parti, Phenaki, Imagen-Video, Text-to-3D, Reinforcement Learning, Optimization, and ML Infrastructure. Before Google, Tris was a Software Engineering Manager at IBM, leading teams to build data science and ML software for retail and consumer products.

Michael Chiang

Co-Founder, Ollama

Michael Chiang is a co‑founder of Ollama, an open‑source platform that lets developers run popular AI models. He also helped create Docker Desktop, which is now used by more than 20 million developers after Docker acquired his first company.

Jedrzej Kosinski

Founding Engineer at ComfyUI

Jedrzej Kosinski is a software engineer who began contributing to the ComfyUI community in the fall of 2023 and has since become a core member of the team.

Dominik Kundel

Developer Experience, OpenAI

Dominik is currently working on Developer Experience & SDKs at OpenAI. Previously he led Product & Design for Twilio's Emerging Tech & Innovation organization. Dominik loves tinkering with anything that can run JavaScript, from front-end servers to CLIs and coffee machines. You can find him tweeting @dkundel and in his spare time he's working on cocktails, food and photography.

Yannik Marek

comfyanonymous, Creator of ComfyUI

Yannik Marek is the creator and cofounder of ComfyUI, maintaining the open-source repository for the better part of three years.

Yineng Zhang

SGLang Core Maintainer

Yineng Zhang is the Principal AI Researcher at Together AI and is a core contributor to FlashInfer. He led critical releases enabling inference for models such as Llama 3 and DeepSeek V3, as well as deployments on GB200 NVL72 and he is also a selected member of LMSYS Org.

Ion Stoica

Professor and Xu Bao Chancellor Chair at UC Berkeley, Director of Sky Computing Lab, and Executive Chairman of Databricks and Anyscale

Ion Stoica is a Professor in the EECS Department and holds the Xu Bao Chancellor Chair at the University of California at Berkeley, the Director of Sky Computing Lab, and the Executive Chairman of Databricks and Anyscale. He is currently doing research on AI systems and cloud computing, and his work includes numerous open-source projects such as SkyPilot, vLLM, ChatBot Arena , Ray and Apache Spark. He is a Member of National Academy of Engineering, an Honorary Member of the Romanian Academy and an ACM Fellow. He also co-founded four companies, LMArena (2025), Anyscale (2019), Databricks (2013) and Conviva (2006).

Robert Shaw

Director of Engineering, Red Hat AI

Robert Shaw serves as the Director of Engineering at Red Hat and has also held the position of Committer at vLLM. With a focus on engineering leadership and contributions to open-source projects, Robert Shaw demonstrates a strong commitment to driving innovation and quality in software development.

Lin Qiao

Co-Founder, CEO, Fireworks AI 

Lin Qiao is the CEO and co-founder of global AI inference cloud and infrastructure platform Fireworks AI, which enables teams like Cursor, Uber, Doordash and Samsung to build, tune, and scale highly-optimized generative AI applications. Prior to founding Fireworks, Lin was the head of Meta’s PyTorch.

Jeff Boudier

VP of Product, Hugging Face

Jeff Boudier builds products at Hugging Face, the #1 open platform for AI builders.
Previously Jeff was a co-founder of Stupeflix, acquired by GoPro, where he served as director of Product Management, Product Marketing, Business Development and Corporate Development.

Linda Yang

Staff. AI Solutions Management, Supermicro

Linda Yang has built her career at the intersection of foundational technologies and innovation. Starting as a software developer for Cisco's IOS, she later played a pivotal role in deploying the largest Smart Grid of its time, serving as the hardware product manager for outdoor ruggedized routers, which also expanded her exposure to edge computing and machine learning.Her growing interest in AI/ML led her to PayPal, where she and her team delivered a unified AI platform for PayPal developers globally—a significant career milestone. She continued to leverage this expertise at ServiceNow, modernizing AI infrastructure from hardware stack to AI services such as training and inferencing.As discussions on sustainable & scalable AI infrastructure intensified, Linda joined Supermicro. She believes in their vision to lead the tech industry towards a more sustainable future and fulfill the environmental promises.

Mark Saroufim

Software Engineer, Meta
Co-Founder, GPU Mode

Mark Saroufim is a Software Engineer at Meta Platforms and core contributor to PyTorch, specializing in GPU optimization and performance engineering. He co-founded GPU MODE, an open-source research community with over 16,000 members dedicated to making GPU programming more accessible.

Workshop (Beginner): Developing Multi-Model Multi-Agent Systems

A beginner to intermediate level instructor-led, hands-on GPU workshop where you will learn how to set up an OpenAI compatible endpoint to serve multiple models concurrently using PyTorch, vLLM, or SGLang and build multi-agent applications to deliver real-world applications. The workshop culminates with a short application development challenge.

Workshop (Beginner): An Inside Look into Captivating Video Generation

A beginner level, instructor-led, hands-on GPU workshop to show you how to generate captivating video clips using popular workflows using open-source tools within PyTorch and ComfyUI. You will learn from industry experts the tips and tricks of setting up a compute efficient pipeline for prompt video workflow, and at the end, have an opportunity to participate in a challenge to win prizes.

Teaching Models to Think: GRPO with Unsloth on a MI300X GPU

Every major AI lab is using reinforcement learning to reach AGI. But why is reinforcement learning important? How does it actually work, and what are RL environments, reward functions, reward hacking, and GRPO? In this talk, we'll explore real-world use cases of reinforcement learning across different domains and see how you can experiment with it yourself, completely locally, by using Unsloth and AMD GPUs.

Workshop (Advanced): How to Develop Optimized Kernels

An advanced user, instructor-led, hands-on GPU workshop with an introduction to AMD GPU architecture and developing optimized kernels. Explore GPU architecture, learn how OpenAI Triton’s block-programming model helps easy development of performant kernels and pick up pro tips for squeezing peak

Additional Talk: From Models to Kernels: Powering AI with AMD GPUs

Jeff Boudier
VP of Product, Hugging Face

In this hands-on lightning talk, learn how to use the latest and greatest open models on Hugging Face on AMD Instinct GPUs with out-of-the-box performance. Then, learn how you can contribute and benefit from AMD AI acceleration with ROCm kernel implementation in Hugging Face libraries.