Own Ricursive's heterogeneous compute infrastructure and build scheduling systems for diverse hardware environments.
Posted by employer 1 day ago
First seen on Joblaze 4 hours ago
Last verified on the company career page 4 hours ago
What you'll build
Must have
Nice to have
Requirements
Not disclosed in this posting: compensation, work arrangement, visa sponsorship.
Joblaze summary
In this role, the engineer manages Ricursive Intelligence's diverse computing infrastructure, which includes GCP GPU pools, Linux workstations, and FPGA boards. Key responsibilities involve designing job scheduling and experiment management systems that efficiently allocate resources across various hardware environments. This position is ideal for a seasoned professional with a strong background in both hardware and software, particularly those who have experience building infrastructure from the ground up. The team is composed of highly skilled individuals from top tech backgrounds, fostering an environment of innovation and expertise.
Joblaze insights
Quick facts
From the original posting
ABOUT THE ROLE
We run our own infrastructure, and it extends well past the cloud. Our engineers depend on GCP compute pools for GPU, GKE, and GCE workloads, on the license servers behind our commercial EDA tools, on a Bazel-based build platform, and on physical hardware: Linux workstations and FPGA boards that our designs run on directly. This role owns that whole estate and the systems that decide what runs where.
You will build job scheduling and experiment management that work the same way whether a job lands on a GPU pool, a build worker, or a board on a bench. That takes someone who is as comfortable debugging a board or an embedded toolchain as designing a distributed system, and who can break a heterogeneous platform into pieces a small team can run. You will work closely with our infrastructure and silicon engineers.
WHAT YOU WILL DO
Own Ricursive's heterogeneous compute: GCP GPU, GKE, and GCE pools alongside Linux workstations and FPGA boards, from bring-up through health monitoring and failure recovery.
Design and build scheduling systems capable of supporting jobs that run on heterogeneous hardware under a wide range of operating conditions and constraints.
Build experiment management systems that enable engineers and researchers to launch, track, and reproduce results across environments.
Own the build platform: Bazel with remote execution on Buildbarn, scaled for our hardware and software repositories.
Keep the license servers for our commercial EDA tools available and fairly shared across teams.
Work hands-on with hardware and embedded tooling as the platform grows, from workstation setup to board-level debugging.
MINIMUM QUALIFICATIONS
BS in CS, CE, EE, or a closely related technical field, or equivalent practical experience.
5+ years building infrastructure or systems that others depend on. [confirm level]
Has worked across the hardware and software boundary, for example hardware bring-up, embedded tooling, firmware, or hardware-in-the-loop infrastructure.
Architected complex systems and decomposed them into components a small team can own.
Built core infrastructure from zero or near-zero, not only operated a mature platform where the infrastructure was abstracted away.
Hands-on experience with job scheduling or resource management across shared compute.
PREFERRED QUALIFICATIONS
Owned a Bazel or Buck build platform, including remote execution using Buildbarn.
FPGA, embedded, or firmware experience, including open hardware projects such as OpenTitan or Caliptra.
Infrastructure work at a trading firm, hardware company, or platform team where engineers own bare metal and software together.
Hands-on experience with GCP and GKE.
Operated GPU clusters.
Early-stage startup experience: built infrastructure from zero or near-zero.
Standard company text repeated across Ricursive Intelligence's postings is omitted here.