CF

Designing Multi-Agent Deep Search Systems Course

1h 47m 55s
English
Paid
Added to CourseFlix

Designing Multi-Agent Deep Search Systems Course is a 1-lesson 1 hour 47 minutes self-paced course by Youssef Hosni. Most AI search engines stop at the stage of searching and retrieving information.

Course facts

Lessons
1
Duration
1 hour 47 minutes
Language
English
Added to CourseFlix
October 2026
Instructor
Youssef Hosni
Price
Premium

Most AI search engines stop at the stage of searching and retrieving information. They receive a user's query, search for information on the internet or within a knowledge base, gather several documents, and generate an answer. For simple queries, this is sufficient, but complex research tasks quickly reveal the limitations of this approach: planning, decomposition of the question, source comparison, relevancy checks, working with contradictions, and structured synthesis of information are required.

This workshop is dedicated to designing multi-agent Deep Search systems that go far beyond ordinary search and basic RAG.

You will learn how to design a Deep Search architecture where agents independently plan research, break down complex questions into sub-tasks, use various tools, gather evidence from multiple sources, evaluate information quality, consider data dates and relevancy, resolve contradictions, and combine results into a reliable final answer.

This is a practical-architectural workshop focused on system design. The main focus is on key components, architectural solutions, trade-offs, and design patterns for building reliable Deep Search agents, without delving deep into coding.

What You Will Learn

During the workshop, you will explore:

  • What Deep Search agents are and why they are needed
  • The difference between LLM, AI agents, and Deep Search agents
  • Why ordinary search and basic RAG are insufficient for complex research tasks
  • How to design the architecture of a multi-agent Deep Search system
  • The roles of Planner, Executor, Validation, and Merger agents, as well as the continuous improvement layer
  • How to break down complex topics into individual research questions
  • How to design iterative search cycles
  • How to determine when to continue searching and when the information is sufficient
  • How to organize agent interaction with tools for search, web scraping, OCR, API, databases, and data enrichment
  • How tools differ from MCP
  • Why high-quality tool descriptions are crucial for reliable agent behavior
  • Why strict data schemas and validation using Pydantic are necessary
  • How to assess the reliability of sources and the quality of obtained information
  • How to work with contradictions between different sources
  • How to consider the relevance of information, dates, and data changes over time
  • How to combine information from multiple sources
  • How to preserve source provenance and confirmation
  • How to form a clean and structured final result
  • How to manage short-term context and long-term memory of agents
  • How to create a continuous improvement layer for the Deep Search system
  • How to conduct a comprehensive evaluation of Deep Search agents from start to finish

What's Included in the Course

  • Full workshop recording — access to the entire session recording, available to review at your own pace.
  • Complete slide set — the presentation used during the workshop, including architectural diagrams, workflows, and design patterns.
  • Technical notes / handbook — a detailed 200-page guide with key concepts, architectural solutions, and practical patterns discussed during the session.
  • GitHub repository with code — source code for creating a Deep Search agent similar to the system discussed in the workshop.
  • Multi-Agent Deep Search Architecture Blueprint — a visual system diagram showing the interaction of planners, executors, validators, result merging layers, memory, and improvement mechanisms.
  • Tool design guide — examples of tools for Deep Search systems, including search, web scraping, OCR and image processing, video-to-text conversion, data enrichment, and reasoning tools.

Who This Workshop Is For

The course will be beneficial to:

  • AI engineers developing agent systems
  • Data Scientists transitioning to LLM application development
  • Developers working with RAG, AI agents, and search systems
  • Technical founders creating AI assistants for research
  • Specialists who want to learn how to design reliable Deep Search systems and multi-agent AI systems

Who teaches Designing Multi-Agent Deep Search Systems Course? Youssef Hosni

Youssef Hosni thumbnail

Youssef Hosni — AI Engineer, Data Scientist, and Applied Scientist, a researcher in the field of machine learning and AI. He is currently affiliated with Aalto University in Finland, where he works on the application of AI agents and ML, as well as the development of AI products. On GitHub, he describes himself as “AI Engineer & Applied Scientist | PhD Researcher @AaltoUniversity”.

In his professional description, it is also noted that he works as a Senior Data Scientist at Ment, where he is involved in implementing Generative AI features into the company's products. Previously, he was engaged in research in the field of Deep Learning and Computer Vision, including applications to medical images.

His specialization aligns well with the courses he is currently offering:

  • Data Science and Machine Learning
  • Generative AI and LLM
  • AI Agents
  • RAG and Agentic RAG
  • Prompt Engineering
  • LangChain
  • Claude Code
  • AI Engineering
  • Python
  • Docker
  • Time Series Analysis
  • Computer Vision
  • technical training and preparation for AI/Data Science interviews

What lessons are included in Designing Multi-Agent Deep Search Systems Course?

This is a demo lesson (10:00 remaining)

Watch the first 10 minutes free. One payment unlocks all 1 lessons here and every other premium course. No auto-renew.

Watch all 1 lesson from $6.67/mo
0:00
/
#1: Building a Multi-Agent Deep Search System
All Course Lessons (1)
#Lesson TitleDurationAccess
1
Building a Multi-Agent Deep Search System Demo
01:47:55
Watch the other 0 lessons

One payment unlocks this course and every other premium course. No auto-renew.

See plans

Books

Read Book Designing Multi-Agent Deep Search Systems Course

#TitleTypeOpen
1Building a Multi-Agent Deep Search System —Unlock
2Slides —Unlock

What courses are similar to Designing Multi-Agent Deep Search Systems Course?

More courses by Youssef Hosni

Frequently asked questions

Should I already know how RAG and AI agents work?
Some familiarity with LLMs, retrieval, and AI agents would help you follow the architectural discussion, but you don't need to assume this is a coding-heavy workshop. You'll look at the differences between an LLM, an AI agent, and a Deep Search agent, as well as where basic RAG falls short. No formal prerequisites are specified, so treat that background as useful preparation rather than a stated entry requirement.
Will I build a working search system, or mainly learn how to design one?
Expect to focus on designing the system rather than coding a complete application. You'll work through how research agents should plan a task, split questions into smaller pieces, gather evidence, and combine their findings. The practical emphasis is on components, design patterns, and architectural trade-offs. A finished, runnable project isn't promised, so this is a better fit if you want to reason through a system before implementing it.
Would this help if I'm trying to improve an existing AI search feature?
It's a relevant fit if your search feature can retrieve documents but struggles with questions that require several research steps. The focus is on what happens beyond retrieval: planning the investigation, comparing sources, checking relevance and dates, and dealing with conflicting evidence. You'll be looking at system-level decisions, so expect more help thinking through the research workflow than tuning an individual search component.
How is this different from a basic RAG tutorial?
A basic RAG workflow typically retrieves information and uses it to produce an answer. Here, you'll focus on research tasks where that single pass isn't enough. The architectural questions include how agents decide what to investigate, divide the work, assess evidence, and reconcile contradictions before producing a structured response. The distinction is about coordinating a research process, not simply adding more retrieved documents to an LLM's context.
Is this a good choice if I mainly want hands-on coding?
Probably not if detailed implementation is your main goal. The workshop explicitly prioritizes system design without going deeply into code. You'll examine how the pieces of a multi-agent research system fit together and what trade-offs those choices create. If you're looking for extensive coding exercises or step-by-step implementation guidance, those aren't established parts of the offering. Choose it primarily for architectural understanding rather than programming practice.
How much time should I set aside?
The listed runtime is 1 hour and 47 minutes, presented as a single workshop lesson. That's the viewing commitment; any time you spend pausing to sketch an architecture or think through your own research use case would be additional. No separate exercise workload is specified. You could set aside roughly two hours to watch it, or allow more room if you want to take notes on the design trade-offs.
What could I apply afterward beyond one particular search project?
The most transferable ideas are ways to organize research and evaluate the evidence it produces. You'll explore breaking a complex question into manageable tasks, assigning work across agents, checking whether information is relevant and current, and combining conflicting findings into a structured answer. Those design concerns aren't tied to a named framework here, so the value is in the reasoning patterns rather than learning one tool's API.