Mohlo by se vám také líbit

Více informací o knize AI Reliability Engineering

• Parametry knihy
• Anotace
• Oblíbené z jiného soudku

Nákupem získáte 82 bodů

Anotace knihy

What happens when an AI system passes every test-and still fails in the real world?
AI Reliability Engineering: Testing, Evaluating, and Deploying AI Systems You Can Trust is a practical guide to building AI systems that are not only powerful, but dependable, measurable, secure, and ready for real-world use.
Modern AI systems behave differently from traditional software. Machine-learning models, large language models, recommendation systems, and AI agents can produce variable outputs, respond differently to changing data and context, and degrade after deployment. Because of this, accuracy alone is not enough to establish trust.
This book explains how to engineer reliability across the entire AI lifecycle through testing, evaluation, data quality, monitoring, deployment controls, security, human oversight, and continuous improvement.
Inside, you will learn how to:

• Test machine-learning models, generative AI applications, and AI agents.
• Detect hallucinations, overconfidence, bias, prediction errors, drift, and hidden failure patterns.
• Evaluate AI behavior across realistic datasets, edge cases, subgroups, and changing conditions.
• Build stronger data pipelines with validation, consistency checks, lineage, and monitoring.
• Assess LLMs for prompt variation, retrieval quality, groundedness, safety, refusal behavior, and guardrails.
• Deploy AI systems safely using release gates, version control, staged rollouts, rollback plans, Docker, and CI/CD.
• Monitor production systems for declining quality, operational instability, changing user behavior, and emerging risk.
• Protect AI applications from prompt injection, adversarial attacks, data leakage, and unauthorized tool use.
• Design meaningful human oversight and risk-management processes.
• Turn production incidents into better tests, controls, alerts, and future releases.

The book also introduces practical frameworks, code examples, checklists, evaluation methods, and real-world case studies across healthcare, finance, fraud detection, customer support, recommendation systems, and autonomous AI agents.
You will see how tools such as Python, pandas, NumPy, scikit-learn, MLflow, Great Expectations, Evidently, Ragas, DeepEval, LangSmith, Docker, and CI/CD systems can support reliable AI engineering.
At its core, this book teaches one essential principle:
A trustworthy AI system is not defined by model performance alone. Reliability depends on the data, software, prompts, retrieval systems, permissions, monitoring, deployment environment, and people surrounding the model.
By the end, you will understand how to decide whether an AI system is ready for production, define acceptable operating limits, identify critical risks, monitor changing conditions, and determine when a system should be restricted, rolled back, improved, or retired.
AI may never be perfectly predictable.
But its uncertainty can be measured, its risks controlled, its failures detected, and its operation supported by evidence.
If you want to move beyond simply building AI that works and start engineering AI systems that can be tested, deployed, monitored, and trusted with greater confidence, this book provides the framework to do it.

Parametry knihy

• Plný název: AI Reliability Engineering
• Podnázev: Testing, Evaluating, and Deploying AI Systems You Can Trust
• Autor: Trent B. Presley
• Jazyk: Angličtina
• Vazba: Brožovaná
• Počet stran: 340
• EAN: 9798192710234
• ID: 53530751
• Nakladatelství: Independently published
• Hmotnost: 791 g
• Rozměry: 280 × 216 × 18 mm
• Datum vydání: 14. August 2026

---

---

---