The MAD Podcast with Matt Turck

Why This Ex-Meta Leader is Rethinking AI Infrastructure | Lin Qiao, CEO, Fireworks AI

Mar 27, 2025
Listen Now

Summary

In this podcast episode, Lin Qiao, the CEO of Fireworks AI, discusses her transition from Meta to leading a new entity focused on simplifying AI infrastructure. The episode covers her leadership role in PyTorch, a pivotal machine learning framework, and the emerging challenges in deploying generative AI technologies. Qiao elaborates on optimizing model performance through managing computational constraints, scaling model components individually, and addressing deployment challenges in varied environments. A significant focus is placed on the accessibility of generative AI for businesses lacking deep expertise, and the evolving landscape of AI infrastructure, including cost reductions in inference. The episode highlights generative AI's transformative role across industries and discusses the importance of developer-centric solutions, paving the way for accelerated innovation. Additionally, Jiao touches upon the implications of transitioning PyTorch to an independent model and its impact on the AI community.

Key Takeaways

  • 1AI infrastructure simplification is key for widespread adoption.
  • 2Cost reductions in AI inference are democratizing access to technology.
  • 3Function calling enables interactivity in AI applications.
  • 4Scaling model components individually can enhance performance.
  • 5Generative AI remains a transformative force across industries.
  • 6Digital natives gain an edge in adopting innovative AI technologies.
  • 7The evolution of AI tool usability is crucial for developer engagement.
  • 8Understanding deployment environments is critical to AI model success.

Notable Quotes

"So if it's compute bound, then we should just add more GPUs to it, right?"

"So by scaling those different parts of model execution independently, then we remove all possible bottlenecks."

"To Fireworks AI, we aim to take away all this complexity from them."

"One of the biggest successes we saw from the PyTorch experience is simplicity scales."

"People don't want to spend time figuring out how to make things work. They just want it to work."

"I believe this overall AI infrastructure would go down by order of magnitude in terms of cost. And they should."

"Now it's happening across the board, everywhere."

"They just want it to work."

"There were tens of those AI frameworks at that time. But none of those were focusing on the simplicity part."

"Currently, we're on top of five different clouds. We plan to expand to possibly 10 different clouds and a lot more regions globally."

"The default mode when we work with enterprises is that no data can leave our premise for good reasons."

"And this whole trend of infrastructure should become a utility is really, really good for the industry."

"There are market demands or constraints that are not movable and there are constraints that are movable."

"We want to take over, solve a big chunk of those complexities and simplify it for the agentic developers in the future."