Principal AI Performance Modeling Architect

Lead GPU architecture specification and performance modeling for large-scale AI training and inference at AMD. Drive 2-3x performance gains through innovative system design and optimization of distributed ML systems. Collaborate with researchers and internal teams to advance AI infrastructure.

• Lead performance modeling and optimization for multi-trillion parameter LLM training/inference including Dense, Mixture of Experts (MoE) with multiple modalities. Model/optimize novel parallelization strategies across tensor, pipeline, context, expert and data parallel dimensions. Architect memory-efficient training systems using structured pruning, quantization, continuous batching, speculative decoding. Incorporate and extend SOTA models like GPT-4 and multi-modal architectures. Collaborate with internal and external stakeholders to disseminate results.

• Extensive experience optimizing large-scale ML systems and GPU architectures. Deep expertise in CUDA programming, GPU memory hierarchies, and hardware-specific optimizations. Proven track record architecting distributed training systems handling large scale systems. Expert knowledge of transformer architectures, attention mechanisms, and model parallelism techniques.

• PyTorch, CUDA, TensorRT, OpenAI Triton. Distributed systems: Ray, Megatron-LM. Performance analysis tools: NSight Compute, nvprof, PyTorch Profiler. KV cache optimization, Flash Attention, Mixture of Experts. High-speed networking: InfiniBand, RDMA, NVLink. MS/PhD in Computer Science/Engineering.