1 month ago
Shanghai, ChinaSenior
Responsibilities
- Develop and optimize AI kernels and operators for Qualcomm Cloud AI platforms.
- Enable and optimize LLM, MoE, and multimodal models for large-scale cloud inference.
- Drive accuracy, performance, efficiency, graph, compiler, kernel, and memory optimizations across the software stack.
- Improve model memory footprint and ROM/RAM efficiency.
- Collaborate with performance and modeling teams to deliver inference efficiency.
- Analyze and resolve performance bottlenecks using profiling, benchmarking, and modeling methodologies.
Requirements
- Master's degree or above in Computer Science, Artificial Intelligence, Mathematics, Electrical Engineering, or a related field.
- At least 5 years of experience in Cloud AI inference, kernel/operator development, or AI compiler technologies.
- Proficiency in C++, Python, and Linux development environments.
- Deep understanding of Transformer architectures, including LLMs, MoE, and modern foundation models.
- Strong experience with MLIR, Triton, TVM, or similar compiler frameworks, including graph transformation, operator lowering, scheduling, operator fusion, kernel auto-tuning, code generation, and memory scheduling.
- Strong experience with parallel computing architectures, vectorization, tiling, memory hierarchy optimization, workload scheduling, graph optimization, model accuracy validation, performance optimization, and memory optimization.
- Preferred experience optimizing Attention, FlashAttention, KV Cache, Embedding, GEMM, Softmax, and MoE routing operators.
- Preferred hands-on experience developing custom operators and high-performance kernels using compiler frameworks or manual techniques on GPUs, NPUs, TPUs, or custom AI hardware.
- Good verbal and written English communication skills for collaboration across global teams.
Categories
About Qualcomm
Qualcomm is a public semiconductor company headquartered in San Diego, founded in 1985, that designs and sells wireless chipsets and platforms for mobile devices, automotive, IoT, and networking, notably the Snapdragon application processors and 5G modems. It also licenses a large portfolio of cellular patents to device makers, generating revenue alongside chip sales; its technology underpins many Android smartphones and emerging automotive and edge-compute systems, and it trades on NASDAQ as QCOM.