Experience with Python for scripting and automation
Familiarity with Docker for containerization
Proficient in GPU programming using CUDA, OpenCL, or ROCm/HIP
In-depth knowledge of GPU microarchitecture and optimization techniques
Responsibilities
Architect and implement high-throughput, low-latency acceleration pipelines using GPUs or FPGAs
Develop host-side APIs, device drivers, and runtime libraries in C/C++
Collaborate with software engineers and researchers to define hardware-software solutions
Benefits
Opportunity to work on cutting-edge government mission programs
Collaborative environment with experts in software engineering and algorithm research
Engagement in high-performance computing and GPU architecture projects
Full Job Description
Active TS/SCI w/ Polygraph required.
Overview: SageCor is seeking a senior software engineer to design and build high-performance acceleration pipelines for a government mission program. The work centers on GPU architecture and development. It spans the full path from accelerator kernels through host-side software. The engineer works closely with software engineers, algorithm researchers, and system architects to define the right hardware-software solution for demanding throughput and latency requirements.
Responsibilities:
Architect and implement high-throughput, low-latency acceleration pipelines using GPUs (CUDA/HIP) or FPGAs (RTL/HLS)
Architect and implement host-side APIs, device drivers, and runtime libraries in C/C++ that stream data between CPUs and accelerators
Partner with software engineers, algorithm researchers, and system architects to define optimal hardware-software solutions
Required Qualifications:
Expert-level C/C++
Python
Docker
CUDA, OpenCL, and ROCm/HIP
Deep understanding of GPU microarchitecture, including warp/wavefront schedulers, shared memory utilization, and occupancy optimization
Preferred Qualifications (FPGA):
SystemVerilog, Verilog, or VHDL
C/C++ for HLS
AMD/Xilinx Vivado or Vitis, or Intel Quartus Prime
Integration of high-speed IP such as PCIe Gen 4/5, DDR4/5, HBM, or 100G Ethernet
