AI is a production workload now. This is the field manual for the IT Administrators who have to keep it running.
Most IT Administrators learned the craft the hard way: self-taught, late at night, one incident at a time. AI arrives as one more incoming tide, bringing new domains of support, new tools, and new expectations, most of them undocumented.
AI & IT: Administrator's Guide translates AI into operator language. It explains what changes in infrastructure, security, reliability, governance, and organizational design when models become both a core workload and a core service. Written for IT Administrators, platform and infrastructure engineers, SRE and operations teams, IT managers, and technical program leaders, it connects first principles to daily decisions: how the AI stack differs from a traditional application stack, how to plan and run AI-capable datacenters and hybrid clouds, and how to govern AI services so they remain safe, predictable, and accountable.
At the center of the book is the AI OSI Model : seven layers and ninety-six sub-layers that map the AI ecosystem the way the original OSI model mapped networking. Its companion roster catalogs more than nine hundred platforms, protocols, tools, and controls, each placed at a specific sub-layer, scored for popularity and maturity, and cross-referenced by the traversals that connect it to the rest of the stack. An administrator can start from a defined operational question and land on the right candidates, rather than comparing the entire AI market at once.
The book is supported by companion online resources maintained by the author at localis.services, including an interactive version of the full roster, a Learning Priority Index that ranks tools by where study time returns the most, and assessment instruments for hyperscaler fit, provider concentration, and compliance burden.
Volume One covers:
The AI OSI Model in full: seven layers, ninety-six sub-layers, and an annotated roster with placement logic, cross-layer traversals, and maturity scoring
AI datacenter and cluster fundamentals: GPUs, accelerators, memory, storage, power, cooling, observability, and upgrade realities
Link-layer and fabric behavior: NICs, DPUs, RDMA, lossless Ethernet, routing domains, congestion, and collective communication for distributed training
Data and storage for AI: ingestion, feature processing, lineage, dataset movement, vector databases, and metadata bottlenecks
Kubernetes and MLOps: training compared with inference, GPU scheduling and sharing, lifecycle gates, promotion toolchains, and rollback discipline
Operations that hold up under pressure: service desk triage, ten families of failure mode, incident response and forensics, reporting and dashboarding
Security for AI systems: identity, access, data protection, supply-chain risk, and threat mapping across all seven layers
Governance, risk, and compliance: control surfaces by steward family, audit readiness, and policy-to-control translation
Organizational interfaces: how IT works with data science, product, security, facilities, procurement, and leadership
This book carries IT leaders from AI anxiety to operational readiness, with frameworks, checklists, metrics, and diagrams for building and operating reliable AI services, whether you run on-premises, in a hyperscaler, or in a hybrid pattern.
