InfiniAI Lab co-designs algorithms, model architectures, and hardware-aware infrastructure to push the frontier of large-scale machine learning — from sparse architectures to agentic RL and real-time multimodal generation.
MonarchRT brings efficient attention to real-time video generation, and Vortex enables efficient and programmable sparse attention serving for AI agents.
Designing and optimizing algorithms on modern hardware to accelerate large machine learning systems.
Beidi Chen
Assistant Professor, Carnegie Mellon University · Amazon Scholar
Beidi Chen is an Assistant Professor at Carnegie Mellon University and Amazon Scholar. Previously, she was a visiting Research Scientist at FAIR, and before that, a postdoctoral scholar at Stanford University. She received her Ph.D. from Rice University.
Her research focuses on efficient AI — specifically, designing and optimizing algorithms on modern hardware to accelerate large machine learning systems. Her work has won the best paper runner-up at ICML 2022 and the best paper award at MICRO 2025. She has been recognized as a Rising Star in EECS by MIT and UIUC, and is a recipient of the Google ML and Systems Junior Faculty Award as well as the Google Research Scholar Award.
InfiniAI Lab develops efficient and scalable AI models and systems by co-designing algorithms, model architectures, and hardware-aware infrastructure.
We build trusted and verifiable AI systems, developing methods to secure models and agents, protect privacy, and verify generated code and system behavior. Our goal is to make AI infrastructure safe, reliable, and robust as models and tools co-evolve.
We prepare infrastructure for a world where AI agents are first-class operators of the AI stack itself — systems that let agents build, deploy, optimize, and evolve ML infrastructure with minimal human intervention.
We bring generative AI into the interactive, multi-sensory loop where humans actually work and create — developing streaming text, image, audio, and video models with sub-second latency, unified multimodal architectures, and human–AI collaboration interfaces that enhance creativity rather than replace it.
We rethink dense computation to unlock long-context and long-generation efficiency, expand model capacity at constant compute through Mixture-of-Experts and conditional computation, and study the scaling laws that govern how sparse models grow with data, parameters, and hardware.
We build the systems stack for the next generation of reinforcement-learned, tool-using agents — large-scale and asynchronous RL pipelines, agentic serving engines, and kernel- and system-level optimizations for sparse attention and MoE inference at frontier scale.
We are currently open for intern, PhD, and postdoc applications.
Send a brief note explaining why you're interested and attach a CV, including your undergraduate grades. No separate cover letter or certificates needed.
Use "Application PhD" or "Application Postdoc" as the email subject. If you're applying to a specific advertisement, note it in the email.
Email beidic@andrew.cmu.eduFill out our application form with your CV and transcript. Then email Prof. Chen and the PhD student(s) or postdoc(s) you'd like to work with, including: (1) CV, (2) transcript, (3) a brief description of your research interests.
Open application formWe're always looking for curious researchers and engineers who want to build the systems behind the next generation of AI. Reach out if our directions resonate.
