Building Agentic AI Workloads – Crash Course

The course offers a solid introduction to agentic AI, balancing theoretical concepts with practical implementation. The instructor’s background in neuroscience and bioinformatics adds credibility, and her industry experience as a machine learning architect ensures relevance. The content is well-organized, progressing logically from history to hands-on coding. The distinction between workflows and agents is clearly explained, and the discussion of architectural patterns provides valuable insights. The course emphasizes practical considerations, such as choosing the right LLM and evaluating agentic systems, which is beneficial for practitioners. However, some claims lack formal citations, and the rapid pace may overwhelm beginners. The inclusion of real-world incidents and career impact adds a pragmatic dimension. Overall, the course is a valuable resource for those seeking to understand and build agentic AI systems, though it assumes some prior knowledge of AI concepts.

The title accurately reflects the content, which is a comprehensive crash course on building agentic AI workloads, covering theory, implementation, and evaluation.

8/10

The course is presented by a machine learning architect with a PhD in neuroscience and bioinformatics, and includes practical Python demonstrations and references to established concepts and protocols. The content is well-structured and aligns with current industry knowledge, though it lacks formal citations for some claims.

• 00:00Introduction and Speaker Background
• 01:15A Brief History of Artificial Intelligence (1940s–Present)
• 05:43Traditional Machine Learning vs. Generative AI
• 06:35The Three Pillars of AI: Algorithms, Data, and Compute
• 11:08Specific Tasks vs. General Task Execution
• 14:41Defining Agency and the Spectrum of Autonomy
• 18:00Agentic Milestone Timeline (2017–2026)
• 20:31What is a Generative AI Agent?
• 23:04Agents vs. Workflows: Dynamic Flow vs. Static Paths
• 26:18Pros and Cons of Agentic Systems
• 29:59Patterns and Anti-patterns: When to Use Agents
• 32:36The Core Components of an Agent
• 34:55Choosing the Right LLM for Your Agent
• 37:38Crafting Identity with System Prompts
• 39:00Understanding Memory: Intrinsic, Short-term, and Long-term
• 41:26Enhancing Capabilities with Tools and Actions
• 43:09Hands-on Implementation: From Single LLM Call to Python Agent
• 52:18Adding Memory and History to Your Custom Agent
• 54:53Building Agents with Frameworks (LangChain)
• 57:17The Evolving Landscape of Models and Frameworks
• 1:00:15Agentic Architectural Patterns: Supervisor vs. Swarm
• 1:01:41Case Study: Single Agent vs. Supervisor Architecture
• 1:04:48Deep Dive: Swarm Architecture Performance
• 1:06:08When to Choose Multi-agent Systems
• 1:09:05Interface Protocols: MCP, A2A, and AGUI
• 1:12:06How to Evaluate Agentic Systems (LLM vs. System vs. App)
• 1:13:53Evaluation Methods: Code-based, LLM-as-a-Judge, and Human
• 1:15:25Current Challenges: Hallucinations, Cost, and Debugging
• 1:18:15Real-world Incidents and the AI Incident Database
• 1:21:28Career Impact: Which Jobs are Most at Risk?
• 1:23:41Software 3.0: The Evolution of Development Paradigms
• 1:29:00Weathering the Storm: Strategies for the Future
• 1:33:40Beyond LLMs: World Models and the Future of AMI
• 1:37:15Recommended Resources and Closing Thoughts

• ML105_Agents GitHub Repository — Slides and labs for the course
• Rola Dali LinkedIn Profile — Instructor's professional profile
• freeCodeCamp News — Platform for related articles and tutorials

• freeCodeCamp — Platform hosting the course and related educational content

• scrimba.com