Principal Software Architect - High Performance Computing

Roles and Responsibility

• As a Software Architect, you will be responsible for design and implementation of robust, scalable infrastructure solutions combining diverse processors (CPUs, GPUs, FPGAs).
• You will analyze and partition workloads to the most appropriate compute unit, ensuring tasks like AI inference and parallel processing runs on specialized accelerators, while serial tasks run on CPUs.
• You will work closely with cross-functional teams, including Algo engineers, product managers, and business stakeholders, to understand requirements and translate them into architectural/software designs that meet business needs.
• You will be coding and developing quick prototypes to establish your design with real code and data.
• You will be a subject Matter expert to unblock software engineers in the HPC domain.
• You will be expected to profile entire cluster of nodes and each node with profilers to understand bottlenecks, optimize workflows and code and processes to improve cost of ownership.
• Conduct performance tuning and capacity planning, monitoring GPU metrics (e.g., using NVIDIA DCGM) for reliability
• Evaluate and recommend appropriate technologies and frameworks to meet project requirements.
• Lead the design and implementation of complex software components and systems.
• Ensure that software systems are scalable, reliable, and maintainable.
• Your primary focus will be on ensuring that the software systems are scalable, reliable, maintainable and cost effective.

Qualifications

• 12 to 18 years of experience in implementing robust, scalable, and secure infrastructure solutions combining diverse processors (CPUs, GPUs, FPGAs)
• Working experience of GPU inference server like Nvidia Triton.
• Very good knowledge C/C++, Data structure and Algorithms and complexity analysis.
• Experience in developing Distributed High Performance Computing software using Parallel programming frameworks like MPI, UCX etc.
• Experience in GPU programming using CUDA, OpenMP, OpenACC, OpenCL etc.
• In depth experience in Multi-threading, Thread Synchronization, Inter process communication, and distributed computing fundamentals.
• Experience in Inter Process communication using Shared memory and Pipes.
• Experience in performance profiling at application and system level (e.g. vtune, Oprofiler, perf, Nividia Nsight etc.)
• Experience in low level code optimization techniques using Vectorization and Intrinsics, cache-aware programming, lock free data structures etc.
• Familiarity with microservices architecture and containerization technologies (docker/singularity) and low latency Message queues.
• Excellent problem-solving and analytical skills.
• Strong communication and collaboration abilities.
• Ability to mentor and coach junior team members.
• Experience in Agile development methodologies.

Additional Qualifications:

• Experience in HPC Job-Scheduling and Cluster Management Software (SLURM, Torque, LSF etc.)
• Good knowledge of Low-latency and high-throughput data transfer technologies (RDMA, RoCE, InfiniBand)
• Good Knowledge of Parallel processing and DAG execution Frameworks like Intel TBB flowgraph, OpenCL/SYCL etc.