Building Production Multimodal AI Agents Vision, Video, Screens, Documents, Retrieval, Memory, and Evaluation | 4.15 MB
Title: Building Production Multimodal AI Agents Vision, Video, Screens, Documents, Retrieval, Memory, and Evaluation
Author: Adrian Morgan Sebuliba
Category: Operating Systems, Internet, Artificial Intelligence
Language: English | 730 Pages |
ISBN: 9781809315458
Description:
Build production-ready multimodal AI agents that reason across text, images, audio, and video. Learn orchestration, tool use, evaluation, observability, security, and deployment patterns for reliable real-world AI applications
Key Features
[*] Build multimodal agents that combine text, vision, speech, video, retrieval, and tool use
[*] Design reliable orchestration, memory, structured outputs, and human-in-the-loop workflows
[*] Evaluate, secure, monitor, and deploy multimodal agents with production-focused practices
Book Description
AI agents are moving beyond text-only conversations to systems that can interpret documents, images, speech, and video and act on the information they extract. Building a convincing demonstration is relatively straightforward; making these agents accurate, safe, observable, and dependable in production is a much harder engineering challenge.
Building Production Multimodal AI Agents provides a practical path from multimodal model capabilities to deployable agent systems. You will learn to design pipelines for text, vision, audio, and video; manage cross-modal context; connect retrieval and external tools; and coordinate agents with explicit state and control boundaries. Practical examples show how to handle uncertain inputs, preserve evidence, structure outputs, and add human review where decisions carry risk. The book then focuses on the production disciplines that separate prototypes from dependable services: representative evaluation datasets, perception and tool-use testing, tracing, security, latency and cost management, recovery from failures, and deployment monitoring.
By the end, you will be equipped to architect and improve multimodal agents for document processing, visual analysis, voice-driven workflows, and other real-world applications.
What you will learn
[*] Select multimodal models and design input-processing pipelines
[*] Process images, documents, audio, video, and mixed-modal context
[*] Build tool-using agents with state, memory, and orchestration
[*] Ground outputs with retrieval and evidence-linked reasoning
[*] Evaluate perception accuracy, task success, and failure modes
[*] Apply permissions, privacy controls, and human approvals
[*] Monitor latency, token usage, cost, and operational quality
[*] Deploy, test, and improve resilient multimodal AI services
Who this book is for
This book is for AI engineers, software developers, ML engineers, solution architects, and MLOps/LLMOps practitioners building AI applications that process multiple data types. It is suited to developers moving beyond text-only chatbots toward agents that interpret images, audio, video, and documents and take controlled actions. Readers should be comfortable with Python, APIs, and basic LLM concepts; prior experience building multimodal agents is not required.
Table of Contents
[*] The Frame Behind "This"
[*] An Evidence Event, Not a Model Answer
[*] Keep Each Modality Hones
[*] Build the First World State
[*] Pixels Are Not Objects
[*] Ground the Claim
[*] Read the Scene
[*] Test What the Agent Sees
[*] Detection, Segmentation, and Tracking Are Evidence Too
[*] An Image Has No Before
[*] Sample What Matters
[*] Find the Moment
[*] Resolve "This" While the World Moves
[*] See the Screen, Read the Interface
[*] Prefer Structure When It Exists
[*] Turn Observations into Actions
[*] Build a Safe Screen Agent
[*] A Page Is More Than Its Text
[*] Search Beyond Text
[*] Cross-Modal Search
[*] Hybrid Retrieval and Reranking
[*] Multimodal RAG with Citations
[*] Memory Is an Event Store
(N.B. Please use the Read Sample option to see further chapters)
DOWNLOAD:
Код:
https://rapidgator.net/file/697a1fe77f198709fc108bdf5d76c0a0/Building_Production_Multimodal_AI_Agents_Vision_Video_Screens_Documents_Retrieval_Memory_and_Evaluation.rar