Erik Bernhardsson, founder of Modal, talks about the challenges of building a serverless platform and creating a new vision for serverless cloud. They discuss the benefits of using Modal for code deployment and setting up GPU-based infrastructure. They also explore the features and use cases of Modal, including its similarities to a CDN and its applications in AI and non-AI tasks. The podcast concludes with a discussion on Modal's future plans and expansion.
Modal Labs is a serverless platform for data teams that abstracts away complexities and enables fast feedback loops.
Modal Labs developed a proprietary file system that optimizes container startup, improving cache efficiency and supporting various use cases.
Deep dives
Modal Labs: A Cloud Platform for Data Teams
Modal Labs is a cloud platform designed specifically for data teams. It helps data teams with tasks like working with machine learning, AI, compute-intensive processes, building data pipelines, and more. Modal allows users to focus on code and abstracts away the complexities of provisioning, resource management, and scaling. By describing code as modal Python, users can easily execute their code in the cloud and pay per usage. Modal aims to make data teams more productive by providing a fast feedback loop and eliminating the need for manual infrastructure setup.
The Journey of Modal Labs: Solving Data Infrastructure Challenges
Eric Bernhardson, CEO of Modal Labs, shares his journey of developing the cloud platform. With a background in coding and previous experience at Spotify, Eric saw the need for a better tool to streamline the production process for data teams. He wanted to build a platform that eliminates the complexities and time-consuming tasks associated with scaling, scheduling, and provisioning. Modal Labs aims to rethink the entire data stack by focusing on the runtime layer and providing a cloud infrastructure that executes code quickly, enables fast feedback loops, and supports the specific needs of data teams.
Innovative Solutions for Fast Container Startup
One of the challenges Modal Labs faced was the need for fast container startup, especially when working with GPU models. Eric and his team built their own file system to optimize the process, leveraging checksum-based content addressable file system architecture. This architecture allows Modal to fetch files from the network on demand, eliminating the need to copy unnecessary data and improving cache efficiency. The proprietary file system delivers low latency container startups and supports a range of use cases, such as stable diffusion, 3D rendering, web scraping, and more.
Expanding Horizons: Beyond AI and Towards Enterprise Adoption
While Modal Labs has found success with AI-related use cases like stable diffusion and GPU-based models, they are also exploring other applications. Eric sees opportunities in biotech, finance, computational biotech, and beyond. As they continue to scale and aim for general availability, they are investing in performance improvements and addressing feedback from users. Modal Labs also plans to attract enterprise customers and recently achieved SOC 2 compliance. Looking ahead, the team is excited about the potential for new features like safe code execution environments and continually pushing the boundaries of what is possible in cloud infrastructure.
This week we talk with Erik Bernhardsson about Modal, a serverless platform for data teams. Erik talks about his background in data science and machine learning, and how he saw a need for a better tool for data teams. We talk about the challenges of building a serverless platform, and how Modal is building a new vision for serverless cloud. We also talk about the challenges of building a platform that is both easy to use and flexible enough to handle a wide variety of use cases.