Research
AI Systems Performance Engineering by Chris Fregly - is it worth it? [D]
*AI Systems Performance Engineering* by Chris Fregly is a ~1,000-page book (published December 2025) covering GPU/CUDA kernel tuning, PyTorch optimization, distributed training, and high-throughput...
AI Systems Performance Engineering by Chris Fregly is a ~1,000-page book (published December 2025) covering GPU/CUDA kernel tuning, PyTorch optimization, distributed training, and high-throughput inference, aimed at engineers seeking actionable strategies to co-optimize hardware, software, and algorithms for both training and inference workloads. The book includes a 200+ item performance checklist and has received broadly positive reception in the ML community. Fregly is a credentialed author with prior O'Reilly titles (Data Science on AWS, Generative AI on AWS) and deep industry experience at AWS, Databricks, and Netflix.
Related
- [[d-60-matmul-performance-bug-in-cublas-on-rtx-5090-d|[D] 60% MatMul Performance Bug in cuBLAS on RTX 5090 [D]]]
- Anyone have an S3-compatible store that actually saturates H100s without the AWS egress tax? [R]
- Started a video series on building an orchestration layer for LLM post-training [P]
- What if your HNSW index stored 3-bit embeddings instead of float32? [R]
Source: research