Crusoe
AI & HPC managed services
Building managed infrastructure for GPU-intensive AI training and inference workloads — Kubernetes orchestration across NVIDIA and AMD accelerators, Slurm scheduling, and the operational tooling around them.
// Senior Staff Software Engineer
Rami AlGhanmi.
AI infrastructure·Distributed systems·Crusoe
I design and operate the cloud-native orchestration platforms that abstract away the underlying infrastructure complexity of GPU-intensive AI workloads. Two decades across platform engineering, distributed systems, and DevOps — taking ownership of the most complex architectural and operational challenges. When platform engineering succeeds, the infrastructure becomes invisible — enabling teams to ship at velocity.
01
What I work on
— focus areas
01 · ai/hpc
AI & HPC infrastructure
Managed services for GPU-intensive training and inference — NVIDIA and AMD accelerators on Kubernetes, Slurm scheduling, and heterogeneous compute at scale.
02 · distributed systems
Distributed systems at scale
Architecting the data, compute, and control planes that run production — and ensuring high availability and resilience under peak loads.
03 · operations
Operating at production scale
Incident response, on-call systems, capacity planning, and observability — the operational discipline that transforms a running system into an enterprise-grade service.
04 · devops
DevOps & GitOps
Self-service CI/CD, infrastructure-as-code, container delivery, and the deployment automation behind two decades of production systems.
02
Selected work
— last decade
Workday
Operational DataLake migration to AWS
Two-phase migration architecture — using AWS DataSync for ingestion and EMR Spark for transformation — that decoupled data transfer from processing logic, enabling thorough downstream validation and reuse. Delivered with zero downtime.
Workday
Unified observability across the fleet
Architected and operated the EKS-based telemetry platform that consolidated a sprawl of legacy tooling into a single pane of glass across all global environments — reducing CapEx and providing high-fidelity operational signals.
Symantec
First cloud-native product to production
Led the production deployment of Symantec Endpoint Protection Cloud — the company's first SaaS product — and built the self-service CI/CD pipeline behind it from scratch.
03
Trajectory
— twenty years, five chapters
| Years | Where | What |
|---|---|---|
| 2025 – now | Crusoe | AI & HPC infrastructure. Managed services for GPU workloads — NVIDIA and AMD accelerators on Kubernetes at scale. Current |
| 2019 – 2024 | Workday | Distributed infrastructure, DevOps tooling, and fleet-wide observability. DataLake migration to AWS. Kubernetes platform for public-cloud delivery with zero-downtime deploys. |
| 2014 – 2019 | Symantec | Cloud security. First SaaS product to production. Established in-house DevOps practice — self-service CI/CD, IaC, and microservice containerization with Docker & Kubernetes. |
| 2008 – 2014 | USC · NASA JPL | MS & PhD coursework. Earth-science data systems at JPL. Built a git-based assignment-delivery and grading pipeline as TA — early DevOps instincts. |
| 2004 – 2008 | KFUPM | BS, Computer Engineering. Hardware-software fundamentals. |
04
About
— in brief
I've spent two decades building infrastructure that has to work — Kubernetes platforms, data pipelines, observability systems, and the operational discipline behind all of them. Currently at Crusoe, building managed AI/HPC infrastructure for GPU workloads. Previously at Workday (distributed systems and DevOps) and Symantec (cloud security).
I care about systems that are operable and sustainable, not just designed. The real challenge lies in the edge cases, migrations, and incident recovery — the crucial details that rarely make it into architecture diagrams. I am patient with detail, focused on resolving engineering friction over workplace politics, and calm on the bridge when leading incident resolution under pressure.
If you want the long-form: a written CV is here.
Stack
kubernetesslurmawsterraformargocdansiblelinuxgopythonhelmprometheussparkhadoop
// elsewhere
Connect.
Open to conversations about AI infrastructure, platform engineering, and hard problems at scale.