AI Performance Engineer
Core
Optimize training and multi-node inference performance across next-gen networking silicon, adapters, switches, and software stacks for AI/HPC datacenters.
Role type
Senior IC AI Performance Engineer
Builds
High-performance distributed AI training and inference systems on CPU/GPU clusters
Domain
AI/ML Infrastructure, High-Performance Computing, Networking
Deliverable
production ML models
Required skills
Distributed training optimization, Multi-node inference tuning, Network fabric tuning (Omni-Path, InfiniBand, Ethernet), Profiling & benchmarking, Compiler optimization, Linux performance tuning
Preferred skills
Cluster scheduling, GPU path profiling, HPC storage systems
Technologies
PyTorch, DeepSpeed, Megatron-LM, RCCL, NCCL, vLLM, llama.cpp, GCC, LLVM, MPI, Nsight Systems, VTune, Prometheus, Grafana, RDMA, NVMe
Responsibilities
Measure and optimize end-to-end performance for distributed AI workloads; Benchmark and tune open-source AI workloads on compute/storage/network hardware; Design and optimize distributed serving topologies and caching strategies; Validate RDMA/GPUDirect RDMA and congestion control; Build proof points comparing interconnects; Instrument and visualize performance metrics; Evangelize best practices through documentation and publications.
Seniority
Senior, hands-on IC