About this role
The team at Modal is building a new infrastructure layer for AI, providing instant GPU access, sub-second container starts, and native storage to support workloads like low-latency inference, fine-tuning, and production sandboxes. Software engineers stepping into the machines team will focus on designing, building, and maintaining high-performance systems for the serverless platform. Daily responsibilities include managing the fleet of bare metal and cloud hosts, developing control planes that provision and monitor hardware, and automating the integration of new capacity from various providers.
Modal exists because AI touches every sector and requires infrastructure that older cloud models cannot support, as evidenced by rapid growth and backing from major venture partners. Engineers who succeed here typically possess deep systems knowledge, a background in operating physical hardware or building control planes, and a pragmatic approach to tackling complex bugs that span from the Linux kernel to Python control-plane services.
What you will do
- Design, build, and maintain the novel, high-performance systems that comprise the serverless platform machines layer.
- Automate the integration of new capacity from a growing set of hardware providers by auditing and benchmarking hosts and clusters.
- Maintain machine images and configure GPUs, RDMA, networking, and storage to bring machines into production.
- Build automation that keeps the fleet healthy without human intervention by detecting issues like bad GPUs, thermals, and disks.
- Investigate problems sitting between hardware and software, such as kernel panics, boot broadcast storms, and container runtime architectures.
- Participate in the on-call rotation to respond to production incidents and troubleshoot complex failures.
What you need
- Must have 5 or more years of experience writing high-quality production code.
- Requires hands-on experience operating physical hardware fleets, including bare metal provisioning, BMC/IPMI, PXE or network boot, and firmware, or building control planes for them.
- Must possess strong cloud skills.
- Requires strong knowledge of low-level operating system foundations such as the Linux kernel, drivers, networking, file systems, and containers.
- Must be effective at debugging across layers ranging from BGP flapping and vBIOS bugs to Python control-plane services.
- Requires a willingness to step into the thick of it with the on-call rotation and respond to production incidents.
Nice to have
- Experience with GPUs and the NVIDIA software stack in production including drivers, health monitoring, XIDs, and RDMA/NVLink
- Prior experience with Go
Skills and tech stack
Questions candidates ask
What kind of hardware and infrastructure will I be working with?
You will work on Modal's machines layer, which consists of a fleet of bare metal and cloud hosts that run functions, sandboxes, and training jobs. The work involves auditing and benchmarking hosts, configuring GPUs, RDMA, networking, and storage, and integrating new CPU, GPU, and storage servers.
What does day-to-day engineering work look like on the machines team?
Engineers design, build, and maintain high-performance serverless systems, automate hardware integration, and build automation that keeps the fleet healthy without human intervention. You will also debug issues across layers, from BGP flapping and vBIOS bugs to Python control-plane services, and participate in an on-call rotation.
What compensation is associated with this role?
The salary for this position is listed as $250K to $300K.