HPC Architect

Were Hiring | Principal Software Architect High-Performance Computing

Location: Chennai | Bengaluru
Experience: 12–18 Years
Level: E4
Employment Type: Full-Time

About the Role:

We are looking for an experienced Software Architect with deep expertise in High-Performance Computing, GPU infrastructure, parallel programming, and heterogeneous computing.
The role involves designing and optimizing high-performance solutions for low-latency, high-throughput image processing and deep-learning workloads, working across CPUs, GPUs, and FPGAs.

Key Responsibilities:

• Design and implement scalable HPC infrastructure using CPUs, GPUs and FPGAs.
  • Architect workload partitioning across different compute units for optimal performance.
  • Develop prototypes and contribute hands-on code to validate architectural designs.
  • Lead performance profiling, optimization and capacity planning across HPC clusters.
  • Optimize GPU workloads and monitor GPU infrastructure using tools such as NVIDIA DCGM.
  • Work closely with algorithm, product and engineering teams to translate requirements into technical solutions.
  • Act as a technical SME and provide architectural guidance to engineering teams.
  • Evaluate technologies and frameworks for scalable, reliable and cost-effective solutions.

Must-Have Skills:

• 10–20 years of experience in HPC / Software Architecture / Heterogeneous Computing.
• Strong programming skills in C/C++ and Data Structures amp; Algorithms.
• Hands-on experience with CUDA / GPU Programming.
• Experience with MPI, UCX or similar parallel programming frameworks.
• Strong understanding of multithreading, synchronization, IPC and distributed computing.
• Experience with GPU inference platforms such as NVIDIA Triton.
• Hands-on experience with performance profiling tools such as VTune, perf, NVIDIA Nsight, etc.
• Experience with low-level optimization, vectorization, intrinsics and cache-aware programming.
• Experience with Docker/Singularity and microservices architecture.
• Strong understanding of parallel and distributed systems.