Ryan Crawford

- Led GB300 new product introduction, serving as the SME for system setup, configuration, validation, - and benchmarking. - Directed NVL72 NVLink Switch product introduction, including imaging, configuration, validation, and production deployment. - Headed XDR InfiniBand and NVIDIA UFM initiatives, driving next-generation interconnect adoption across large-scale GPU clusters. - Spearheaded HPC engineering efforts for the first GA deployment of B200 GPU clusters, establishing scalable architectures optimized for large AI/ML workloads. - Contributed to the company’s first Managed Slurm offering, enabling automated workload orchestration and hybrid-cloud HPC operations. - Led validation and installation of the largest HPC cluster deployment in 2024 for a major enterprise customer, ensuring on-time delivery and high performance—resulting in multiple follow-on system deployments. - Served as InfiniBand SME for cloud-scale HPC environments supporting 10,000+ GPUs. - Pioneered NVIDIA SHARP implementation across cloud-based HPC clusters to accelerate collective communication performance. - Led end-to-end deployment of 6+ GPU clusters and played a major role in 20+ cluster installations, including on-prem DGX systems for healthcare institutions, government agencies, and academic research organizations. - Defined and validated cluster software and firmware stacks, including OS (RHEL/CentOS), NVIDIA drivers, CUDA, OFED, InfiniBand HCA firmware, switch firmware, and system BIOS. - Designed and executed cluster validation and stress-testing frameworks across 12+ deployments using tools such as NCCL, NVBandwidth, FIO, DCGMI, GPU_burn, and MLPerf benchmarks. - Drove engineering initiatives to integrate third-party cloud HPC solutions, improving interoperability and scalability across hybrid infrastructure.

- HPC Team lead for 2 DOD customers, overseeing 6 engineers maintaining 4 systems, 2 of which have landed in the top half of the Top500 - Primary Slurm technical expert for DOD sites first Slurm clusters - Installed 2 Nvidia DGX’s into a commercial vendor environment, installed all hardware and software, including the configuring additional 2 infiniband switches. - Successfully led team through 3 major DOD system security audits - Planned and mentored team through RHEL 8.8 OS upgrades on all clusters - Headed planning,installment, and integration of an additional 80 node rack into 1500 node cluster. - Headed benchmarks of new installation, including build config of benchmarks such as GAMESS, HYCOM, LINPAK, MPI Ping Pong, and Storage Benchmarks MDTEST and IOR - Installed and configured Robinhood for Lustre, for advanced storage metrics - Assisted in the application of Hot Pool functionality on Lustre file system, allowing for flash storage and hard drives to be in same file system and data to be moved from to the other when utilized or created. - Maintaining and configuring multiple versions of openmpi ranging from 4.1.6 to 5.0.1 with use for aocc,gcc,intel, nvhpc compilers

- Lead On-Site Engineer for install, implementation, and maintenance of DOD HPC Cluster (8.2 petaFLOPS, 176,128 compute cores, 382 TB of memory, and 26 petabytes of storage) 1 - Configured and Installed Mellanox InfiniBand high speed network across 157 IB switches and 4000+ links - Administered a DragonFly Plus topology through UFM - including SHARP implementation - Headed First HPCMP Program & Site Slurm implementation, advanced QoS’s, plugins, accounting. Integration with program Software License server and User Management system. - Primary POC and Liaison between Customer and Vendor relation - Led Contract Acceptance Testing in the minimum possible window of Completion. Completing System Effectiveness Level Testing and System Capabilities Testing in 34 days with an expected completion time of 60 days. - Assisted in the install of 3 Cooling Distribution Units and 37 MotivAir cooling doors - Primary resource for hardware failure diagnosis and replacement - includes AMD CPU’s, Gigabyte Motherboards, NVIDIA A100’s and A40’s, Mellanox switches, Bel Power PDU’s, Lustre Controllers and SFA’s. - Utilized Ansible and Clusterware for node imaging and cluster configuration - Oversaw 2 RHEL upgrades from versions 8.4 to 8.5 to 8.7 clusterwide. - Lead engineer responsible for strict system requirements including 98% uptime, 75%+ of system utilization, minimal system interrupts and node down time, over 30 day periods. - Trained and Led 1 Jr. Linux Admin for first time administration and maintenance of HPC Cluster

- Developed and managed HPC Log Management Tool, used to diagnose day to day critical hardware and software issues - Authored HPC System Report Application, identifying file misconfigurations, Lustre status, Infiniband and Fabric Issues - Primary military liason and tech lead for 2 DOD HPC systems - Utilization of CMU - Cluster management utility - Configured and installed Ganglia monitoring tool across multiple MPP clusters - Accomplished Nagios and Thruk script automation checks - Experienced in configuring and troubleshooting SLURM - Familiar with Configuration Management tools such as BitBucket, Salt, Git - Experienced in in Imaging nodes using cluster software - Prepared and Instructed 8 week Python Training program for 12 suboordinates - Supervised and Managed 2 airman, mentored through their Career Progression certifying them in over 262 job tasks