Agent Engineering

The core of modern AI engineering. Build agents from first principles.

Цена: 5 896 ₽

Длительность: 43 ч

Автор: John Jackson

Программа курса

  1. The Agent Loop: Observe, Think, Act
  2. ReWOO and Plan-and-Execute: Decoupled Planning
  3. Reflexion: Verbal Reinforcement Learning
  4. Tree of Thoughts and LATS: Deliberate Search
  5. Self-Refine and CRITIC: Iterative Output Improvement
  6. Tool Use and Function Calling
  7. Agent Memory — Virtual Context and Memory Paging
  8. Memory Blocks and Sleep-Time Compute
  9. Hybrid Memory: Vector + Graph + KV
  10. Skill Libraries and Lifelong Learning (Voyager)
  11. Planning with HTN and Evolutionary Search
  12. Anthropic's Workflow Patterns: Simple Over Complex
  13. Stateful Graph Orchestration — Durable Execution and Checkpoints
  14. The Actor Model for Agents — Async Messages and Typed Runtimes
  15. Role-Based Agent Teams — Roles, Tasks, Processes
  16. OpenAI Agents SDK: Handoffs, Guardrails, Tracing
  17. The Harness as a Library — Subagents and Session Store
  18. Production Agent Runtimes — Fast Instantiation and Typed Workflows
  19. Benchmarks: SWE-bench, GAIA, AgentBench
  20. Benchmarks: WebArena and OSWorld
  21. Computer Use: Claude, OpenAI CUA, Gemini
  22. Voice Agents: Pipecat and LiveKit
  23. OpenTelemetry GenAI Semantic Conventions
  24. Agent Observability: Langfuse, Phoenix, Opik
  25. Multi-Agent Debate and Collaboration
  26. Failure Modes: Why Agents Break
  27. Prompt Injection and the PVE Defense
  28. Orchestration Patterns: Supervisor, Swarm, Hierarchical
  29. Production Runtimes: Queue, Event, Cron
  30. Eval-Driven Agent Development
  31. Agent Workbench Engineering: Why Capable Models Still Fail
  32. The Minimal Agent Workbench
  33. Agent Instructions as Executable Constraints
  34. Repo Memory and Durable State
  35. Initialization Scripts for Agents
  36. Scope Contracts and Task Boundaries
  37. Runtime Feedback Loops
  38. Verification Gates
  39. Reviewer Agent: Separate Builder from Marker
  40. Multi-Session Handoff
  41. The Workbench on a Real Repo
  42. Capstone: Ship a Reusable Agent Workbench Pack
  43. Frame the Task Before the Agent Writes Code
  44. Build an Evidence-Backed Execution Plan
  45. Delegate Agent Work with Isolation and Merge Contracts
  46. Turn Every Agent Correction into a System Improvement
  47. Define the Outcome Before You Choose the Output
  48. Discover the Workflow People Actually Perform
  49. Map Assumptions and Resolve the Riskiest One First
  50. Choose the Smallest Slice That Can Change the Decision
  51. Write Specifications That Preserve Judgment
  52. Design Success Metrics Before the Result Exists
  53. Choose Prototype, Pilot, or Production Deliberately
  54. Build a Feedback Ratchet with Ownership and Retirement
  55. Итоговое задание