Kernel Engineer

HCL Technologies Limited; Bengaluru; On-site; INR 4,000,000 - 7,000,000; Full time

Job summary

HCL Technologies Limited is seeking a Senior Kernel Engineer to optimize CUDA and Triton kernels for high-performance training and inference workloads. You will write low-level code close to the hardware and collaborate with runtime teams to ship kernel-level improvements.

Candidates should have 8+ years in systems or GPU performance engineering, deep CUDA/Triton experience, and strong C++ skills. A B.Tech or M.Tech in CS/EE/Physics from a Tier-1 institution is expected.

Qualifications

• Write and optimize CUDA/Triton kernels for training and inference workloads.
• Profile GPU workloads and eliminate memory-bandwidth and compute bottlenecks.
• Implement quantization and mixed-precision kernels (FP8/INT4) to cut inference cost.
• Strong C++ and systems-level, performance-first mindset.

Responsibilities

• Write and optimize CUDA/Triton kernels for training and inference workloads across our systems and models.
• Profile GPU workloads and eliminate memory-bandwidth and compute bottlenecks that generic frameworks leave on the table.
• Implement quantization and mixed-precision kernels (FP8/INT4) that materially cut inference cost.
• Work directly with runtime engineering teams to get kernel-level wins into production quickly.
• Track new GPU architectures as they ship and adapt our kernels to take advantage of them.

Skills

CUDA kernels

Triton

C++

GPU architecture

Performance profiling

Memory optimization

Education

B.Tech or M.Tech in CS/EE/Physics

Job description

Every millisecond of idle GPU time costs us money, and you're the person who gets rid of it. As a Kernel Engineer, you'll write and optimize the low-level CUDA and Triton kernels that determine how fast our training and inference actually run — the kind of close-to-the-metal work that most companies simply can't hire for, because there aren't many people who do it well.

What You'll Do

• Write and optimize CUDA/Triton kernels for training and inference workloads across our systems and models
• Profile GPU workloads and eliminate memory-bandwidth and compute bottlenecks that generic frameworks leave on the table
• Implement quantization and mixed-precision kernels (FP8/INT4) that materially cut inference cost
• Work directly with runtime engineering teams to get kernel-level wins into production quickly
• Track new GPU architectures as they ship and adapt our kernels to take advantage of them

What You'll Bring

• 8+ years of experience in systems, GPU, or performance engineering
• Strong, hands-on experience writing CUDA or Triton kernels
• Solid understanding of GPU microarchitecture and memory-bandwidth-bound optimization
• Proficiency in C++ and a systems-level, performance-first mindset
• Comfort profiling and debugging at the level of individual GPU instructions

Education

B.Tech or M.Tech in Computer Science, Electrical Engineering, or Physics, from a Tier-1 institution.

Nice to Have

• A background at a chip company, EDA vendor, or GPU/accelerator software team — this transfers unusually well to this role even without direct ML experience
• Exposure to quantization or low-precision inference work