Senior CUDA Engineer

Tekskills Inc.; Kolkata District; On-site; INR 2,600,000 - 3,800,000; Full time

Job summary

Tekskills Inc. is seeking a Senior CUDA Engineer in Lucknow, UP to develop and optimize GPU kernels using CUDA, CuBLAS/CuSPARSE, and NVIDIA libraries.

The role emphasizes high-performance matrix operations, memory layout optimization, and seamless Python integration via PyBind11 and PyTorch extensions. The candidate should have 6–8 years of hands-on experience in GPU-accelerated computation, with a focus on deep learning inference, mixed-precision stability, and robust error handling.

Qualifications

• 6–8 years of experience in GPU-accelerated computation and deep learning inference
• Develop and optimize high-performance CUDA kernels for large-scale workloads
• Strong experience with NVIDIA libraries and integration with Python-based orchestration

Responsibilities

• Develop and optimize GPU compute kernels using CUDA, CuBLAS/CuSPARSE, and related libraries
• Design efficient flows for large-scale GEMM/SpMM workloads
• Profile and tune performance with Nsight Compute and Nsight Systems
• Build clean CC APIs and integrate with Python (PyBind11) and PyTorch extensions
• Ensure numerical correctness and robust error handling

Skills

CUDA

C++

GPU kernels

NVIDIA libraries

Tools

cuBLAS

cuSPARSE

CUTLASS

Nsight Compute

Nsight Systems

PyBind11

PyTorch extensions

Job description

Job Title: Senior CUDA Engineer

Location: Lucknow, UP

Skills Required: C++
Experience: 6 to 8 Years

Notice Period: Immediate and 15 Days

Job Description:

• Develop and optimize GPU compute kernels using CUDA, Tensor Cores, and NVIDIA libraries Implement and tune high-performance matrix operations using cuBLAScuBLASLt, CUTLASS, and cuSPARSE
• Design batch grouped execution flows for large-scale GEMM SpMM workloads Optimize memory layouts, data movement, and kernel launch efficiency
• Conduct profiling and performance tuning using Nsight Compute Nsight Systems
• Build clean and stable CC APIs and integrate with Python (PyBind11 PyTorch extensions)
• Ensure numerical correctness, mixed-precision stability, and robust error handling

Essential Skills:

• We are seeking a highly skilled Senior CCUDA Engineer with strong experience in GPU-accelerated computation and deep learning inference.
• The role involves developing high-performance GPU kernels, optimizing large-scale matrix operations, and integrating with Python-based orchestration layers.