1 day ago
Hsinchu, Taiwan or Taipei, TaiwanSenior / Staff+
Responsibilities
- Convert, optimize, and deploy models for efficient inference using PyTorch and ONNX.
- Analyze and optimize the performance of LLM, VLM, and diffusion models for throughput and latency constraints.
- Map current and next-generation AI workloads onto hardware designs using sharding and parallelism strategies.
- Collaborate with customers and internal compiler, firmware, and platform teams to drive AI solutions.
- Analyze complex performance and stability issues to identify root causes.
- Create engineering solutions that provide continuous insights into AI workload performance.
- Design and implement high-level kernels, including Triton kernels, that generate efficient low-level code.
Requirements
- Hands-on experience building and optimizing language models with PyTorch and ONNX, preferably in production environments.
- Deep understanding of transformer architectures, attention mechanisms, and performance trade-offs.
- Strong Python programming skills and experience with workload mapping, sharding, or parallelism.
- Understanding of computer architecture, ML accelerators, in-memory processing, and distributed systems.
- Bachelor’s degree in Computer Science, Engineering, Information Systems, or a related field plus 4+ years of relevant experience; or a master’s degree plus 3+ years; or a PhD plus 2+ years.
- Knowledge of neural-network operators, linear algebra, mathematical operations, math libraries, machine learning compilers, model accuracy evaluation, torch.compile, or torchDynamo is preferred.
- Strong communication, problem-solving, learning, and collaboration skills.
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.