Microsoft
Senior Software Engineer
$120K–$261K
United States · On-site · Senior · Full-time · Posted Oct 26, 2026
- Experience
- Required: 4+ years of technical engineering experience coding. Preferred alternatives: 6+ years with a master's degree, or 8+ years with a bachelor's degree, or equivalent experience.
- Education
- Bachelor's Degree in Computer Science or related technical field, or equivalent experience.
- Visa sponsorship
- The role requires Microsoft, customer, and/or government security screening, including the Microsoft Cloud Background Check.
About the job
Optimizes and troubleshoots network performance for large-scale distributed HPC and AI workloads running on InfiniBand-connected GPU clusters.
What you'll do
- Diagnose communication bottlenecks and analyze InfiniBand fabric behavior, including congestion and link errors.
- Tune and debug RDMA transport, network policies, and topology-aware communication; develop tools to analyze fabric telemetry and traffic.
- Benchmark HPC and AI workloads and collaborate with operations teams and hardware vendors on fabric features and interconnects.
What they ask for
- Bachelor's degree in Computer Science or a related technical field and 4+ years of technical engineering experience coding in a listed language, or equivalent experience.
- Ability to meet Microsoft, customer, and/or government security screening requirements, including the Microsoft Cloud Background Check.
Nice to have
- Master's degree plus 6+ years of technical engineering experience, or bachelor's degree plus 8+ years, or equivalent experience.
- Experience debugging network performance at scale; expertise with MPI, NCCL, InfiniBand/RDMA, GPU communication stacks, and AI workload performance tuning.
Pay
USD $119,800 - $234,700 per year (base salary); USD $160,200 - $261,000 per year (base salary)
Tools
C, C++, C#, Java, JavaScript, Python, MPI, NCCL
AI focus
Optimizes networking and collective communication performance for large-scale AI workloads, including transformer training and model/data-parallel workloads.
Model-written summary · the original posting is authoritative