Principal / Director Software Engineer – AI Accelerators (Datacenter) at Qualcomm

Cork, [object Object]

on-site · full_time · Visa sponsorship

Apply for this role at Qualcomm

Responsibilities:
- Architect, deliver, and own scalable, high-performance AI model serving platforms for datacenter and cloud providers.
- Lead design and development of software stacks for NPUs/accelerators, focusing on LLM inference and fine-tuning workloads.
- Drive runtime and inference optimizations (dynamic batching, scheduling, memory management, latency reduction over RoCE/low-latency fabrics).
- Grow and mentor a mid-sized engineering team (20+), establish best practices for performance, reliability, and observability.
- Collaborate cross-functionally with hardware, research, platform, and product teams to align roadmaps and deliveries.
- Design production-grade distributed systems: orchestration, deployment, high availability, fault tolerance, and monitoring.
- Implement Linux/system-level optimizations and low-level integrations for custom accelerator runtimes.

Requirements:
- 15+ years software engineering experience with 5–8+ years in technical leadership roles.
- Expert-level C/C++ and Python; strong system-level programming experience.
- Hands-on experience with PyTorch and modern ML frameworks and building LLM-serving platforms.
- Practical experience with accelerator execution runtimes (CUDA/ROCm or equivalent), NPU programming, and Linux optimization.
- Proven knowledge of Kubernetes, Triton Inference Server, and LLM serving frameworks (SGLang, vLLM, etc.).
- Deep understanding of distributed systems, high-performance computing, networking (RoCE), and production reliability.
- Experience operating in matrixed, geographically distributed organizations and owning technical roadmaps.