Graphics Compute & Kernel Optimization Engineer

Saaki Argus and Averil Consulting; Bengaluru Urban; In loco; INR 3.000.000 - 6.000.000; Tempo pieno

Descrizione del lavoro

Saaki Argus and Averil Consulting seeks a Senior Graphics Compute Engineer to optimize compute kernels for graphics and neural rendering workloads on high-performance accelerators. You will work with C/C++, GPU frameworks, and compiler toolchains to push the limits of parallel execution.

The role requires deep knowledge of memory hierarchies, synchronization mechanics, and performance benchmarking across CUDA, OpenCL, and Vulkan Compute, with emphasis on NeRF and neural graphics techniques.

Competenze

• Proficiency in C/C++ and GPU programming.
• Experience with parallel computing on multi- and many-core architectures.
• Familiarity with compiler frameworks (LLVM/MLIR).
• Solid understanding of graphics pipelines (rasterization, ray tracing, texturing, framebuffers).
• Low-level kernel development and performance tuning.

Mansioni

• Analyze and adapt graphics and neural workloads (NeRF, 3D splatting) for accelerator architectures.
• Develop and optimize compute kernels for ray marching, volume rendering, image processing, and neural-graphics tasks.
• Collaborate with compiler and hardware teams to improve code generation and memory performance.
• Benchmark, profile, and optimize compute workloads; identify bottlenecks and propose architecture-aware improvements.

Conoscenze

C/C++

GPU programming

CUDA/OpenCL/Vulkan Compute

LLVM/MLIR

Graphics pipeline

Kernel development

Memory optimization

Neural rendering

Formazione

BTech/MTech/MS/PhD in CS/EE/Graphics or related

Strumenti

CUDA

OpenCL

Vulkan Compute

LLVM

Descrizione del lavoro

• Senior Graphics Compute & Kernel Optimization Engineer
• Experience: 8–12 Years
• Education: B.Tech / M.Tech / MS / Ph.D. in Computer Science, Electrical Engineering, Computer Graphics, or a related discipline.
• Role Overview
• We are looking for an experienced Graphics Compute Engineer with strong expertise in low-level programming, parallel computing, compiler technologies, and performance optimization.
• The role involves adapting and optimizing advanced graphics, neural rendering, image processing, and scientific visualization workloads for high-performance, compute-oriented accelerator architectures. The engineer will be responsible for developing compute-based solutions to replicate graphics functionalities that may not be directly supported by the underlying hardware.
• The ideal candidate should have hands-on experience in C/C++, GPU programming, compiler frameworks, memory optimization, and neural graphics technologies.
• Key Responsibilities
• 1. Graphics & Neural Computing Development
• Analyze and adapt advanced graphics workloads, including neural rendering, NeRF, 3D Gaussian Splatting, image reconstruction, denoising, and scientific visualization.
• Develop and optimize compute kernels for ray marching, volume rendering, image processing, and neural network-based graphics applications.
• Implement graphics-related workloads on parallel compute architectures using available software development kits and low-level programming frameworks.
• Explore alternative compute-based approaches to implement graphics functionalities that are not natively supported by the hardware.
• Develop, port, and optimize high-performance kernels using C/C++ and GPU programming frameworks.
• Collaborate with compiler engineering teams to improve architecture-specific code generation using LLVM and MLIR.
• Optimize memory access, data movement, synchronization, and parallel execution across many-core processor architectures.
• Work on SIMD/MIMD execution models and improve workload distribution and computational efficiency.
• Develop software-based alternatives for traditional graphics pipeline operations such as rasterization, texture sampling, and ray tracing.
• 3. Performance Engineering & Benchmarking
• Design and execute performance benchmarks for graphics, rendering, and visualization workloads.
• Evaluate compute performance against established technologies and frameworks such as CUDA, OptiX, and Vulkan Compute.
• Identify performance bottlenecks related to memory bandwidth, compute utilization, synchronization, and kernel execution.
• Analyze hardware limitations and recommend optimization strategies to improve overall workload efficiency.
• Assess the feasibility, development effort, and performance trade-offs involved in implementing graphics functionalities through compute-based approaches.
• 4. Cross-Functional Collaboration
• Work closely with compiler, runtime, architecture, and hardware engineering teams to improve accelerator performance.
• Support the development and integration of optimized compute libraries and runtime components.
• Evaluate emerging neural graphics techniques and their applicability to next-generation compute architectures.
• Required Technical Skills
• Strong programming expertise in C and C++.
• Hands-on experience with GPU programming frameworks such as CUDA, OpenCL, or Vulkan Compute.
• Strong understanding of parallel computing, multi-core and many-core processor architectures.
• Experience with compiler technologies such as LLVM and MLIR.
• Good knowledge of graphics pipeline concepts, including rasterization, ray tracing, texture sampling, and framebuffer operations.
• Experience in low-level kernel development and performance optimization.
• Strong understanding of memory hierarchy, memory management, synchronization, and parallel execution.
• Exposure to neural rendering and graphics technologies such as NeRF, Gaussian Splatting, neural denoising, and super-resolution.
• Preferred Qualifications
• Experience adapting or porting graphics and visualization applications to non-traditional GPU or compute accelerator platforms.
• Exposure to MIMD-based architectures and architecture-specific code optimization.
• Experience with scientific visualization, image reconstruction, rendering engines, or neural graphics pipelines.
• Ability to evaluate hardware capabilities and develop software-driven alternatives for unsupported graphics operations.