Uploaded July 2025 | Updated September 2026, 1 week ago
Learn more about @Scale here: atscaleconference.com
Registered for our latest event, @Scale: Networking: bit.ly/4luYOSF
In episode 2, Joe Spisak, Product Director for Artificial Intelligence at Meta, sat down with Ion Stoica to discuss a wide range of topics related to the AI and machine learning industry, including the resurgence of reinforcement learning, large language models, the power of open source software, and the evolution of the AI tech stack.
Ion Stoica is a Professor in the EECS Department at the University of California at Berkeley, and the Director of SkyLab (https://sky.cs.berkeley.edu/). He is currently doing research on cloud computing and AI systems. Past work includes Ray, Apache Spark, Apache Mesos, Tachyon, Chord DHT, and Dynamic Packet State (DPS). He is an Honorary Member of the Romanian Academy, an ACM Fellow and has received numerous awards, including the Mark Weiser Award (2019), SIGOPS Hall of Fame Award (2015), and several "Test of Time" awards. He also co-founded three companies, Anyscale (2019), Databricks (2013) and Conviva (2006).
Joe Spisak is Product Director and Head of Open Source in Meta’s Generative AI organization. A veteran of the AI space with over 10 years experience, Joe led product teams at Meta/Facebook, Google and Amazon where he focused on open source AI, open science and building developer tools such as PyTorch to help the community scale up AI in an open and collaborative way. As the leader of product for PyTorch, he and the team built an amazing platform and community that made PyTorch the leading open source AI development framework in the industry and took it to the Linux Foundation where it now resides in partnership with Microsoft, Nvidia, AMD, Google, Amazon and more. Joe is also an angel and advisor to companies like Anthropic, Answer.ai, Lastmile.ai, Evolutionary Scale, Udacity, Lightning.AI, and others.
Key Points
- Joe and Ion discussed the rise of reinforcement learning and how it has gained significant momentum in recent years.
- They talked about the importance of large language model evaluations and the work being done in this area, including the development of projects like Vicuna.
- The conversation touched on the open-source nature of the AI and machine learning community, and how this has enabled rapid progress and collaboration.
- Joe and Ion highlighted the challenges of managing and evolving large open-source projects, such as maintaining community engagement, handling breaking changes, and balancing flexibility with stability.
- They discussed the concept of a "stack" for AI and machine learning, including projects like PyTorch, Kubernetes, VLM, and Ray, and how these components are being adopted and integrated by companies.
Main Arguments
- Joe and Ion discussed the rapid growth and adoption of reinforcement learning in recent years. (00:54)
- They highlighted the importance of large language model evaluations and the work being done in this area, such as the development of Vicuna. (01:15)
- They emphasized the open-source nature of the AI and machine learning community, and how this has enabled rapid progress and collaboration. (06:08)
- They discussed the challenges of managing and evolving large open-source projects, such as maintaining community engagement, handling breaking changes, and balancing flexibility with stability. (21:01)
- Joe and Ion talked about the concept of a "stack" for AI and machine learning, including projects like PyTorch, Kubernetes, VLM, and Ray, and how these components are being adopted and integrated by companies. (41:39)
Supporting Evidence
- Joe and Ion provided examples of the growth of projects like Spark, Ray, and VLM, highlighting the rapid adoption and evolution of these technologies. (11:03)
- They discussed their personal experiences in managing open-source projects, such as the challenges of refactoring VLM and the importance of community engagement. (21:01)
- They mentioned their involvement in projects like PyTorch and TensorBoard, and the collaborative nature of these efforts with industry partners. (26:42)
- They provided insights into the dynamics of the AI and machine learning industry, including the shift in compensation and the challenges of convincing students to pursue academic careers. (33:33)
Learn more about @Scale here: atscaleconference.com
Registered for our latest event, @Scale: Networking: bit.ly/4luYOSF
In episode 2, Joe Spisak, Product Director for Artificial Intelligence at Meta, sat down with Ion Stoica to discuss a wide range of topics related to the AI and machine learning industry, including the resurgence of reinforcement learning, large language models, the power of open source software, and the evolution of the AI tech stack.
Ion Stoica is a Professor in the EECS Department at the University of California at Berkeley, and the Director of SkyLab (https://sky.cs.berkeley.edu/). He is currently doing research on cloud computing and AI systems. Past work includes Ray, Apache Spark, Apache Mesos, Tachyon, Chord DHT, and Dynamic Packet State (DPS). He is an Honorary Member of the Romanian Academy, an ACM Fellow and has received numerous awards, including the Mark Weiser Award (2019), SIGOPS Hall of Fame Award (2015), and several "Test of Time" awards. He also co-founded three companies, Anyscale (2019), Databricks (2013) and Conviva (2006).
Joe Spisak is Product Director and Head of Open Source in Meta’s Generative AI organization. A veteran of the AI space with over 10 years experience, Joe led product teams at Meta/Facebook, Google and Amazon where he focused on open source AI, open science and building developer tools such as PyTorch to help the community scale up AI in an open and collaborative way. As the leader of product for PyTorch, he and the team built an amazing platform and community that made PyTorch the leading open source AI development framework in the industry and took it to the Linux Foundation where it now resides in partnership with Microsoft, Nvidia, AMD, Google, Amazon and more. Joe is also an angel and advisor to companies like Anthropic, Answer.ai, Lastmile.ai, Evolutionary Scale, Udacity, Lightning.AI, and others.
Key Points
- Joe and Ion discussed the rise of reinforcement learning and how it has gained significant momentum in recent years.
- They talked about the importance of large language model evaluations and the work being done in this area, including the development of projects like Vicuna.
- The conversation touched on the open-source nature of the AI and machine learning community, and how this has enabled rapid progress and collaboration.
- Joe and Ion highlighted the challenges of managing and evolving large open-source projects, such as maintaining community engagement, handling breaking changes, and balancing flexibility with stability.
- They discussed the concept of a "stack" for AI and machine learning, including projects like PyTorch, Kubernetes, VLM, and Ray, and how these components are being adopted and integrated by companies.
Main Arguments
- Joe and Ion discussed the rapid growth and adoption of reinforcement learning in recent years. (00:54)
- They highlighted the importance of large language model evaluations and the work being done in this area, such as the development of Vicuna. (01:15)
- They emphasized the open-source nature of the AI and machine learning community, and how this has enabled rapid progress and collaboration. (06:08)
- They discussed the challenges of managing and evolving large open-source projects, such as maintaining community engagement, handling breaking changes, and balancing flexibility with stability. (21:01)
- Joe and Ion talked about the concept of a "stack" for AI and machine learning, including projects like PyTorch, Kubernetes, VLM, and Ray, and how these components are being adopted and integrated by companies. (41:39)
Supporting Evidence
- Joe and Ion provided examples of the growth of projects like Spark, Ray, and VLM, highlighting the rapid adoption and evolution of these technologies. (11:03)
- They discussed their personal experiences in managing open-source projects, such as the challenges of refactoring VLM and the importance of community engagement. (21:01)
- They mentioned their involvement in projects like PyTorch and TensorBoard, and the collaborative nature of these efforts with industry partners. (26:42)
- They provided insights into the dynamics of the AI and machine learning industry, including the shift in compensation and the challenges of convincing students to pursue academic careers. (33:33)
