At a glance
Build and operate the machines layer powering Modal’s serverless AI platform, automating bare-metal provisioning, GPU infrastructure, fleet health, and hardware integration at scale.
Summarized by AI from the original posting
What you'll do
- Design, build, and maintain high-performance systems for the serverless platform
- Automate integration of new hardware capacity
- Audit and benchmark hosts and clusters
- Maintain machine images and configure GPUs, RDMA, networking, and storage
- Build automation to detect and remediate unhealthy machines
- Debug hardware, kernel, networking, and container-runtime issues
- Develop automatic hardware acceptance testing and benchmarking
- Manage network bootloaders, kernel, and firmware across the fleet
Requirements
- 015+ years of experience writing high-quality production code
- 02Experience operating physical hardware fleets or building control planes
- 03Strong cloud skills
- 04Strong knowledge of low-level operating system foundations
- 05Effective cross-layer debugging
- 06Willingness to participate in on-call rotation
Skills
Full description
About Us:
AI needs a new infrastructure layer. We're building it at Modal.
Every era of computing brought new workloads that previous infrastructure couldn't support: mainframes, databases, and the cloud. Each time, the company that rebuilt the layer underneath defined the decade. AI is no different, except it touches everything instead of one slice, and the window to build the layer underneath it is open right now.
Our customers include category-defining companies like Lovable, Ramp, Cognition, DoorDash, and Suno. They rely on Modal for instant GPU access, sub-second container starts, and native storage, so it's simple to serve low-latency inference, fine-tune models, and access production-ready sandboxes at scale.
We recently raised a $355M Series C at a $4.65B valuation, led by General Catalyst and Redpoint Ventures. We've crossed $300M+ ARR and grown fivefold since September.
Our team includes creators of popular open-source projects (e.g.,Seaborn,Luigi), academic researchers, international olympiad medalists, and experienced engineering and product leaders with decades of experience.
The Role:
We are looking for strong engineers with experience and interest in designing, building, and maintaining the novel, high-performance systems that make up our serverless platform. Specifically, you'll be working on Modal's machines layer: the fleet of bare metal and cloud hosts that every Function, Sandbox, and training job runs on, and the control plane that provisions, images, monitors, and repairs them. You'll automate the integration of new capacity from a growing set of hardware providers; from auditing and benchmarking hosts and clusters, to maintaining our machine images, configuring GPUs, RDMA, networking, and storage, and getting machines into production. You'll build the automation that keeps the fleet healthy without human intervention: detecting bad GPUs, thermals, and disks. You'll dig into whatever is between the hardware and the software that runs on top of it, whether that is a kernel panic, a broadcast storm during boot, or getting our container runtime to run on new architectures and platforms.
Requirements:
5+ years of experience writing high-quality production code
Experience operating fleets of physical hardware (bare metal provisioning, BMC/IPMI, PXE or network boot, firmware) or building the control planes that manage them (the more challenges you've worked through, the better)
Strong cloud skills
Strong knowledge of low-level operating system foundations (Linux kernel, drivers, networking, file systems, containers, etc.)
Effective at debugging across layers, from BGP flapping, Linux RPS, and vBIOS bugs to a Python control-plane service
Willingness to step into the thick of it with our on-call rotation and respond to production incidents
Nice-to-Haves:
Experience with GPUs and the NVIDIA software stack in production (drivers, health monitoring, XIDs, RDMA/NVLink)
Prior experience with Go
Key Things the Team Is Working On:
Automatic remediation of unhealthy machines (power cycling, reimaging, GPU recovery) to maximize uptime and minimize operator toil.
Automatic integration of new CPU, GPU, and storage servers into the fleet while managing hardware and network heterogeneity.
Network health monitoring and reliability across many datacenters, and standardization of bare metal network configuration.
Automatic hardware acceptance testing and benchmarking (CPU, disk, GPU, interconnect, network).
Custom network bootloader, machine image pipeline, and kernel and firmware management across the fleet.