AI Infra Summit Speakers

Joe Spisak is a Product Director in Meta’s Superintelligence Labs (MSL), where he leads product efforts across PyTorch and Meta’s Agentic platform. He has over a decade of experience building AI platforms, with leadership roles at Meta, Google, and Amazon. Joe helped make PyTorch the world’s leading open-source AI framework and guided its transition to the Linux Foundation in partnership with industry leaders. He spearheaded the open-source strategy for Llama, making cutting-edge models broadly accessible and enabling the community to scale AI in an open and collaborative way. Joe is also an active angel and advisor to next-generation AI startups such as Anthropic, General Reasoning, ReflectionAI and many others.

Tianqi Chen is currently an Assistant Professor at the Machine Learning Department and Computer Science Department of Carnegie Mellon University. He is a Distinguished Engineer at NVDIIA. He received his PhD. from the Paul G. Allen School of Computer Science & Engineering at the University of Washington. He has created many major learning systems that are widely adopted: XGBoost, TVM, and MLC-LLM

Bowen Peng is currently the Chief Scientist at Nous Research, where he leads fundamental research on transformers and generative models. With a Master’s degree from Mila, his research spans distributed training, transformers, and multimodal generative models. He is the lead author of YaRN (Yet Another RoPE Extension Method), a widely adopted innovation in long context large language models and DeMo (Decoupled Momentum Optimization). Bowen’s work focuses on rethinking architectures for multimodal data, large scale distributed training and exploring new paradigms that push beyond today’s transformer-dominated landscape.

I helped fix bugs in Llama, Mistral, Gemma, Phi, Qwen, Mistral and more. We have over 10 million monthly downloads, and collaborate with large model labs to release more accurate and bug free models. Our Github package has >40K stars and is used by NASA, the UN and many others for RL, GRPO, finetuning, inference and more!

Robert Nishihara is one of the co-founders of Anyscale and one of the co-creators of Ray, the leading open source framework for scalable AI. He did his PhD in machine learning and distributed systems in the computer science department at UC Berkeley. Before that, he majored in math at Harvard.

Currently, I’m a PhD student at Berkeley Sky Computing Lab for machine learning system and cloud infrastructures. I am advised by Prof. Joseph Gonzalez and Prof. Ion Stoica. My latest focus is building an end to end stack for LLM inference on your own infrastructure:

• vLLM runs LLM inference efficiently.

Previous exploration includes:

• Conex: builds, push, and pull containers fast.
• SkyATC: orchestrate LLMs in multi-cloud and scaling them to zero.

I previously work on Model Serving System @anyscale.

• Ray takes your Python code and scale it to thousands of cores.
• Ray Serve empowers data scientists to own their end-to-end inference APIs.

Before Anyscale, I was a undergraduate researcher @ucbrise.

I am a member of technical staff at xAI. I lead the inference team responsible for building efficient and reliable infrastructure to run Grok models. My expertise includes machine learning systems, large language models, compilers, and distributed systems.

Previously, I completed my Ph.D. at UC Berkeley, where I was advised by Ion Stoica and Joseph E. Gonzalez. I obtained my B.S. degree from ACM Honored Class, Shanghai Jiao Tong University. I was honored to receive the Meta PhD Fellowship and the a16z Open Source AI Grant (twice) in recognition of my innovative research and impactful open-source projects. I also co-founded the non-profit organization LMSYS.org to advance open research on large models. We have developed open models with millions of downloads, crowdsourced platforms with millions of users, and systems that are orders of magnitude faster.

As a Staff Machine Learning Scientist at Niantic (Scopely), Andy builds scalable ML solutions for one of the biggest mobile games (Pokemon GO) to enhance player experiences and drive business impact. He is passionate about connecting data with product strategy to deliver measurable value.

Nathan is a machine learning researcher who works on building, understanding, and advocating for open language models and other responsible autonomous systems.

He is currently a post-training lead at the Allen Institute for AI.

Mark Saroufim is a PyTorch Engineer at Meta working on inference, compilers and community.

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

My research is in AI systems. I focus on expanding the Pareto frontier between quality and efficiency, to unlock new AI capabilities, by considering AI algorithms, hardware, and applications in lockstep. Recently, I’ve looked at:

• Algorithms: How do we build AI architectures that scale efficiently?
  (Zoology, Based, JRT, LoLCATS, Monarch Mixer)
• Hardware: As the complexity of AI hardware increases, how can we make it easier to extract high utilization for new AI algorithms?
  (ThunderKittens, Megakernels, KernelBench)
• Applications: How do we bridge the efficiency gaps that emerge as models are deployed in new settings, such as data management?
  (AMA for personal data, Cartridges, Evaporate, Data Wrangling)

Will Brown is a Research Lead at Prime Intellect, where he works on open-source research and infrastructure for agentic reinforcement learning, and is the creator of the verifiers library. Previously, he held research roles at Morgan Stanley and Amazon Web Services, completed his PhD in Computer Science at Columbia University advised by Christos Papadimitriou and Tim Roughgarden, and completed his undergraduate studies in Computer Science and Philosophy at the University of Pennsylvania.

Mostafa is a seasoned engineering leader in high-performance computing. During his tenure at NVIDIA, he served as Engineering Director and led teams to develop optimized deep learning libraries like cuDNN and CUTLASS, revolutionizing GPU-accelerated AI. At SiFive, as VP of Software, Mostafa assumed a leadership role guiding teams in the development of an MLIR/LLVM-based software stack for SiFive Intelligence & performance cores. His contributions also extend to optimizing Intel GPU hardware/software features, playing a pivotal role in developing the AVX1/2 SIMD ISA for Intel CPUs, and contributing to the GNU C Compiler. Outside of his professional pursuits, Mostafa enjoys exploring the great outdoors. In his free time, you can often find him hiking in nature, hitting the slopes for a skiing adventure, cycling through scenic routes, or shooting hoops on the basketball court. Mostafa holds a Master of Science in Electrical Engineering from the Technion, with a specialization in compiler optimizations. His unwavering passion for innovation continues to drive advancements in the field of high-performance computing.