AI Algorithms for SREDevOps Reliability and Delivery in the Agentic Era

the gap between ML research papers and operational reality, providing SREs, DevOps engineers,
and platform teams with the algorithmic knowledge needed to operate AI systems reliably.
What you'll learn:
- Size GPU clusters for inference and training workloads
- Debug performance issues in vector search and RAG systems
- Implement rate limiting that works for token-based APIs
- Design resilient AI systems with proper fallback strategies
- Optimize costs without sacrificing quality
- Monitor AI systems for failures that traditional observability misses
Each chapter follows a pattern: here is the algorithm, here is why you care, here is what
to watch, here is what goes wrong. Algorithms are presented as numbered steps with formulas
and decision tables rather than implementation code-the focus is on understanding, not
copying and pasting.
This book is for infrastructure engineers responsible for AI systems: SREs inheriting
LLM-powered applications, DevOps engineers building ML platforms, backend engineers adding
AI features, platform engineers designing GPU infrastructure, and technical leads evaluating
AI infrastructure decisions.
Topics covered include:
- Vector search algorithms (HNSW, IVF, PQ) and their operational implications
- RAG pipeline optimization and retrieval strategies
- LLM inference optimization: KV cache, PagedAttention, continuous batching
- GPU scheduling with MIG, time-slicing, and MPS
- Distributed training: data parallelism, ZeRO, pipeline parallelism
- AI-specific monitoring and drift detection
- Autoscaling strategies for GPU workloads
- Resilience patterns for AI systems
- Cost optimization techniques

Uploady
8d1fc.7z
ClicknUpload
8d1fc.7z
Rapidgator
8d1fc.7z.html
AlfaFile
8d1fc.7z

KatFile
8d1fc.7z.html