Location: Baltimore, MD
Cap-Exempt H-1B Position — No lottery required
Baltimore, MD, United States Find out how well you match with this job Job description Working at JHU Affiliated Office Address Baltimore, MD, United States Requisition ID 121121 Date Created April 30, 2026 Job Family Information Systems, IT Job Subfamily Systems Engineering Job Function Information Technology Exempt Status Exempt Shift Type Full Time Schedule Mon-Fri, 8:30am-5pm Worksite 02-MD:Mount Washington Campus Work Modality Remote: Not on-site or < 10% of hours worked on-site IT@JH Research Computing is seeking a ***HPC Scientific Software Engineer*** to support faculty, researchers, and students engaged in high-performance and AI-driven research across Johns Hopkins University. The position is responsible for deploying, optimizing, and maintaining scientific software and computational workflows on advanced HPC Systems and related infrastructure. Working primarily within Linux-based environments, the engineer manages and troubleshoots complex software stacks, containerized applications, and GPU-accelerated workloads using tools such as SLURM, Easy build, Spack, etc. The role combines ticket-based user support with long-term project work, collaborating closely with interdisciplinary research groups to enhance system performance, streamline data-intensive workflows, and integrate cutting-edge technologies. The position operates with significant independence while coordinating regularly with systems engineers and research computing leadership to ensure reliable, high-efficiency computing resources that advance the university’s scientific mission. **Specific Duties & Responsibilities** *Software Deployment and Design (15%)* - Develop and refine deployment strategies for scientific software on HPC and AI systems. - Design computational workflows, selecting optimal software configurations, and utilizing tools like Ansible for automation. - Assist teams in implementing, tuning, and optimizing AI models and gateway applications (e.g., XDMoD, Coldfront, Open OnDemand