Building AI Solutions with Agentic Frameworks

What's inside

1. Level 1: What an Agent Solution IsFree
   Learn that an agent solution delivers a business capability, not a demonstration, and what that implies for design.
2. Level 2: Purpose and Scope of the ProgrammeFree
   Understand the vendor-neutral purpose of the material and the path it defines from idea to production.
3. Level 3: Extracting Facts as a Discrete Step
   Study fact extraction as the first isolated step in an agent workflow and why isolation matters.
4. Level 4: Agent Patterns: The Core Vocabulary
   Survey the agent patterns named in the document — prompt chaining, routing and parallelisation — and when each applies.
5. Level 5: Prompt Chaining: Structure
   Learn prompt chaining as an ordered sequence of steps, each consuming the prior step's output.
6. Level 6: Prompt Chaining: Draft and Review Steps
   Implement the drafting and review stages of a chain with separated responsibilities.
7. Level 7: Prompt Chaining: Revision and Formatting
   Complete the chain with a revision step driven by review findings and a final formatting step.
8. Level 8: Routing: Classify Then Dispatch
   Use the routing pattern to classify an input and dispatch it to the correct specialised handler.
9. Level 9: Parallelisation: Splitting Work
   Apply parallelisation to run independent subtasks concurrently and aggregate their results.
10. Level 10: Composing Patterns into a Workflow
    Combine chaining, routing and parallelisation into a single coherent agent workflow.
11. Level 11: Phase 1 Review: Patterns Checkpoint
    Consolidate agent solution definitions and the three core patterns through applied scenarios.
12. Level 12: Tools as Agent Capabilities
    Introduce tools as the mechanism by which agents act beyond text generation.
13. Level 13: Reusable Tool Patterns
    Design tools for reuse across multiple agents and solutions.
14. Level 14: Tool Errors, Limits and Safe Failure
    Handle tool failure, timeouts and invalid arguments so agents degrade predictably.
15. Level 15: Framework Selection: Workload Shape
    Learn to characterise workload shape as the primary input to framework selection.
16. Level 16: The Framework Decision Matrix
    Read and apply the framework decision matrix as described in the material.
17. Level 17: Evaluating Framework Fit in Practice
    Validate a shortlisted framework against real constraints before committing.
18. Level 18: Avoiding Framework Lock-In
    Structure solution code so framework choice remains reversible.
19. Level 19: Phase 2 Review: Tools and Frameworks
    Consolidate tool design and framework selection by producing a defensible selection for a scenario.
20. Level 20: Model Choice as a Governed Decision
    Begin governed model access by treating model choice as an explicit, reviewable decision.
21. Level 21: Model Abstraction Layers
    Learn how a model abstraction layer decouples agent code from specific model providers.
22. Level 22: Model Gateways and Governed Access
    Understand the model gateway as the enforcement point for governed model access.
23. Level 23: Gateway Operations: Quotas, Costs and Failover
    Configure gateway-level controls for cost, rate limits and provider failover.
24. Level 24: Building Evaluation Datasets
    Construct representative evaluation datasets that cover expected and edge-case agent behaviour.
25. Level 25: Deterministic Tests for Agents
    Write deterministic tests that assert verifiable agent and tool behaviour without model variance.
26. Level 26: Model-Based Evaluation
    Apply model-based evaluation to score outputs where deterministic assertions are impossible.
27. Level 27: Grounding Evaluation in Practice
    Evaluate whether agent outputs are grounded in the supplied evidence rather than fabricated.
28. Level 28: Red-Team and Adversarial Testing
    Run structured adversarial testing against prompts, tools and agent autonomy boundaries.
29. Level 29: Architecture Views: An Overview
    Learn the four architecture views used to describe an agent solution completely.
30. Level 30: Runtime and Execution Infrastructure
    Design the runtime view covering where agent steps and tools actually execute.
31. Level 31: Runtime Choices for Agent Workloads
    Select a runtime based on workload shape, latency and state requirements.
32. Level 32: Delivery and Supply Chain
    Secure the delivery view, including dependencies, prompts and tool provenance.
33. Level 33: CI/CD for Agent Solutions
    Build a pipeline that gates releases on deterministic, model-based and grounding evaluations.
34. Level 34: Checkpoints and Durable Agent State
    Use checkpoints so long-running agent runs can resume without repeating work.
35. Level 35: Idempotency and Safe Retries
    Make agent and tool actions idempotent so retries cannot cause duplicate side effects.
36. Level 36: Post-Deployment Verification
    Verify an agent release in production using telemetry, tracing and operational metrics.
37. Level 37: Constituent Parts of an Agent Platform
    Assemble the shared agent platform from its constituent parts and critical distinctions.
38. Level 38: Identity for Agents and Tools
    Establish workload and user identity so agent actions are attributable and least-privileged.
39. Level 39: Security and Auditability
    Implement security controls and an audit trail that reconstructs any agent decision.
40. Level 40: Human Control and Escalation
    Design human-in-the-loop approval, override and escalation for high-impact agent actions.
41. Level 41: Exam Readiness: Critical Distinctions Review
    Consolidate the course by rehearsing the answer guide's critical distinctions across patterns, frameworks, governed model access, evaluation and operations.
42. Level 42: Timed Mock Exam (50 questions, 90 minutes) Timed mock
    A timed 50-question, 90-minute mock exam testing the whole course: agent patterns, tool and framework selection, governed model access, evaluation, architecture views, delivery, operations and agent platform governance.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →