Module 1:
Foundations & The Context Packet
-
The Physics of LLMs: Tokenization, Latency vs. Cost, and the Autoregressive nature of generation.
-
Context Engineering: Moving beyond "Prompt Engineering" to the engineering discipline of the "Context Packet"—controlling the Perception-Reasoning-Action loop.
-
Lab: The Reliability Challenge – Iteratively refining a context packet to achieve 95% accuracy under strict token budgets.
Module 2:
The Twelve-Factor Agentic SDLC
-
The Orchestrator Mindset: Transitioning from solo coder to strategic orchestrator.
-
The Testing Pyramid for AI: Introducing Deterministic Anchors – How to separate and test deterministic logic vs. probabilistic AI responses.
-
The Playbook: Writing a "Mission Brief" and "Technical Plan" before writing code.
-
Triage: Strategic decision-making: When to use Async (AI) vs. Sync (Human) logic.
Module 3:
Automation with Spec-Kit
-
Automating the Playbook: Using Tikal’s spec-kit CLI to automate the Agentic SDLC workflow.
-
Spec-Driven Development: The principle of "Debug the spec, not the code."
-
Lab: A fast-paced workshop building a feature from intent to implementation using the CLI.
Module 4:
The Core Engine – ReAct & MCP
-
Agent Protocols: Deep dive into the ReAct loop (Think → Act → Observe).
-
Model Context Protocol (MCP): Architecting the agent as a "Universal Client" that connects to external databases and tools via strict schemas.
-
The Decision Matrix: Engineering judgment: When to use dynamic LLM tools vs. deterministic hard-coded functions.
Module 5:
Self-Healing Architectures & Agentic RAG
-
Beyond Naive RAG: Solving "Lost in the Middle" and data hallucinations.
-
The Self-Correction Loop: Implementing Reflection and CRAG (Corrective RAG) patterns.
-
Memory Systems: Architecting Short-term (Graph State) vs. Long-term (Vector) memory.
-
Lab: "The Self-Doubting Bot" – An agent that critiques its own answers and rewrites queries until confidence is met.
Module 6:
Multi-Agent Orchestration
-
From Tools to Peers: Moving from function calling to Agent-to-Agent (A2A) collaboration.
-
The Agent Card: Standardizing discovery protocols between agents.
-
Organizational Design: Implementing Hierarchical (Supervisor) vs. Swarm (Panel of Experts) architectures using LangGraph.
Module 7:
Production-Ready Agents (Governance & Economics)
-
The Economics of Model Selection: Creating a strategy for Model Routing—dynamically routing simple tasks to cheaper models and complex reasoning to SOTA models to optimize TCO.
-
Testing Non-Deterministic Systems: Building a "Golden Dataset" and running regression tests to detect quality drift.
-
Safety Layers: Implementing Human-in-the-Loop (HITL) gates and deterministic guardrails.
Module 8:
Capstone Project – "The Simulated Reality"
-
The Challenge: Teams design, build, and validate a Multi-Agent System using the Tikal Agent Scaffold.
Mandatory Requirements: -
Self-Healing Logic: The system must detect and fix its own errors.
-
The Evaluation Gate: Must pass a "Golden Dataset" of adversarial inputs.
-
Externalized State: Production-grade state management (Redis/Postgres).
-
Presentation: Teams present their architecture, cost analysis, and "Evolve Loop" story.
