Junior GPU Software Engineer
The Junior GPU Software Engineer at Confidential Client will develop, test, and optimize GPU software components within the OxPython stack, focusing on kernels, runtime, and framework integration for AI workloads. This early-career role involves close collaboration with senior engineers to enhance performance, enable new models, and support software/hardware validation on cutting-edge GPU IP. The position offers a unique opportunity to build deep GPU software expertise while contributing to a production-ready AI silicon platform in an on-site setting.
About Careertakes
👉 Important disclosure: Careertakes is a third-party recruiting platform supporting this hiring process. If selected, you will be employed directly by our client, Semiconductor Manufacturing and Technology.
Applicants for this role may also receive access to additional matched opportunities through the Careertakes platform.
About the Role
Confidential Client is hiring an early-career Junior GPU Software Engineer to join a growing GPU software team focused on building and productionizing a full GPU software stack for AI workloads. This on-site role in Campbell, CA is ideal for engineers with 1–4 years of GPU, systems, or performance-oriented software experience who want to learn from senior engineers and work across kernels, runtimes and framework integration.
What You’ll Do
- Develop, test and optimize GPU software in the company’s stack across kernel, runtime and framework integration layers.
- Implement and tune GPU kernels (CUDA, Triton, custom ops) for AI workloads such as GEMM, attention, convolution and memory-bound operators.
- Profile and benchmark workloads; identify bottlenecks and help close the gap to peak hardware utilization.
- Apply SIMT and parallel compute concepts: thread/warp scheduling, divergence, occupancy and efficient memory-hierarchy usage.
- Support bring-up on pre-silicon and emulation platforms and help enable customer models and operators.
- Collaborate with compiler engineers on lowering kernels via MLIR/LLVM and mapping to hardware.
- Conduct software/hardware validation through tests, debugging and issue triage.
- Perform root-cause analysis for functional and performance issues in kernels and runtime code.
- Contribute to tooling, automation and documentation to improve developer productivity and release readiness.
- Participate in design and code reviews and learn from senior engineers across teams.
Qualifications
- 1–4 years of experience in GPU, systems or performance-focused software (including internships, research or graduate projects).
- Strong fundamentals in C/C++ and systems programming.
- Hands-on GPU programming experience (CUDA) from coursework, research, internships or projects.
- Understanding of SIMT and parallel compute fundamentals: threads/warps, divergence, synchronization, occupancy and memory hierarchies.
- Interest in performance engineering and basic experience with profiling and optimizing code.
- Familiarity with computer architecture and operating systems concepts.
- Exposure to ML frameworks (PyTorch, JAX, ONNX) and how deep learning workloads run on GPUs.
- Strong problem-solving and debugging skills, eagerness to learn, and a bias toward shipping: prototype, measure and iterate quickly.
Preferred (Nice to Have)
- Experience writing or optimizing GPU kernels (GEMM, convolution, reductions, attention).
- Exposure to Triton or similar high-level GPU kernel frameworks.
- Familiarity with compiler infrastructure (LLVM, MLIR).
- Experience with GPU profiling tools (Nsight or similar).
- Experience with AI-assisted development workflows.
- Coursework, research or projects in parallel computing, GPU architecture, compilers or HPC.
- Prior work with LLM inference/training, KV-cache optimization, distributed/multi-GPU execution or serving frameworks (e.g., vLLM).
- Contributions to open-source projects such as LLVM, MLIR, Triton or PyTorch.
Education
- BS, MS or PhD in Computer Science, Computer Engineering, Electrical Engineering or a related field.
Why Confidential Client
- Join a fast-growing team delivering a complete GPU hardware + software IP platform for AI.
- Work directly with hardware architects and senior engineers across the stack.
- Opportunity to grow deep GPU software skills and influence a production platform used by customers.
- Collaborative, hands-on environment where early-career engineers learn quickly and contribute broadly.
Compensation & Benefits
Confidential Client offers a competitive compensation package that typically includes base salary, equity participation and comprehensive medical, dental and vision coverage. Specifics will be provided to candidates during the hiring process.
Equal Opportunity & Hiring Transparency
Careertakes and our client are Equal Opportunity Employers committed to building a diverse and inclusive workforce. We prohibit discrimination or harassment of any kind. To support a fair and efficient hiring process, AI tools may be used to assist with application review or resume screening. These tools do not replace human decision-making. Final hiring decisions are made by people.
If you have questions about how your data is used, please contact us directly.
Negotiate a higher salary! Check the salary ranges for this job type in your area.
View My Salary Range