01—15
Agentic AI
Engineering
Models, tools, reasoning, memory, orchestration, interfaces and production architecture.
Engineering Autonomous Systems
BUILD AGENTS.BUILD WITHAGENTS.
Agentic AI Engineering & Agentic SDLC. Learn to engineer intelligent agents, then manage sophisticated software projects built by teams of agents.
The progression
Two connected disciplines.
Part I teaches how agents work. Part II teaches how to run sustained engineering projects through agents without losing control of scope, decisions or quality.
Agentic AI
16—19
Agentic SDLC
Specifications, parallel agent execution, engineering control, decision governance and verified delivery.
Part I / Sessions 01–15
Agentic AI Engineering.
The technical foundation: from transformer internals to an integrated agent architecture.
| 01 | Transformer Foundations | Attention, embeddings, positional encoding and transformer architecture. |
| 02 | Modern LLM Internals | Pretraining, fine-tuning, alignment, reasoning and multimodal models. |
| 03 | Developer Foundations & First Agent | Python, Node.js, frontend fundamentals and building a tool-using agent. |
| 04 | Model Context Protocol (MCP) | Clients, servers, tools, resources and integration. |
| 05 | Reasoning, Planning & Cognitive Architecture | ReAct, task decomposition and perception–decision–action loops. |
| 06 | Decision Models & Fast Agent Control New | Jev, System 1/2, typed decisions, confidence and LLM fallback. |
| 07 | Memory Systems & Modern RAG | Semantic and episodic memory, hybrid retrieval and persistence. |
| 08 | Multi-Agent Systems & DAG Architecture | Agent collaboration, DAG execution, parallelism and checkpoints. |
| 09 | Browser Agents | Playwright, CDP, web interaction and autonomous research. |
| 10 | Computer Use & Desktop Agents | Screen perception, accessibility and desktop automation. |
| 11 | Channels, Gateways & Event-Driven Agents | Messaging, voice, webhooks, event buses and proactive agents. |
| 12 | Agent Reliability, Security & Isolation | Sandboxing, permissions, retries, recovery and fault tolerance. |
| 13 | Agent-to-Agent Protocol (A2A) | Discovery, delegation and cross-system interoperability. |
| 14 | Agentic User Interfaces (A2UI / AG-UI) | Generative UI, event streaming and user interaction. |
| 15 | Model Routing, Economics & Production Integration | Routing, cost optimization, observability and Arcturus integration. |
By Session 15, students have the foundations to design and build complete agent systems.
Part II / Sessions 16–19
Agentic SDLC.
One continuous engineering project. Four sessions focused on managing software complexity with agents.
| 16 | Agentic Project Architecture & Planning | Requirements, specifications, architecture, milestones, deliverables, ownership, dependency graphs, Gantt planning and acceptance criteria. |
| 17 | Multi-Agent Engineering & Execution | Coding agents, specialized workstreams, Claude Code and Codex, Git worktrees, parallel development, context persistence and handoffs. |
| 18 | Agentic SDLC Control Plane & Decision Governance | Project dashboards, agent reports, locked decisions, approval scope, “Needs You” queues, blockers, progress tracking and replanning. |
| 19 | Evidence-Driven Verification & Production Delivery | Independent review, automated tests, measured evidence, CI/CD, integration, qualification, release management and change control. |
A closer look at Agentic SDLC
Keep the project in control.
Agents can write features quickly. The harder problem is keeping a large project coherent when many agents are working, requirements change, decisions accumulate and test results disagree. Agentic SDLC treats the project itself as a managed engineering system.
Plan & dependencies
A living schedule of deliverables, owners, dependencies, blockers and revised completion forecasts.
Agent workstreams
Who owns each area, what changed, which code was produced, and what remains unfinished.
Decision history
Approved choices stay recorded with their rationale, boundaries, authority and later amendments.
Needs You
Human attention is focused on architectural trade-offs, exceptions, approvals and external dependencies.
Evidence
Test results, reviews and deployment records determine whether work is implemented, measured or fully qualified.
Recovery
A new agent should be able to understand the project state and resume without reconstructing past chat history.
An Agentic SDLC control plane, in miniature
Sample project / Release 01
Illustrative project stateEngineering workstreams
| API contract | Approved |
| Agent execution engine | In progress |
| Frontend integration | Blocked |
| Regression suite | Testing |
Needs You / Decisions
| Confirm schema change | Decision |
| Review integration conflict | Review |
| Authorize release | Pending |
| Decision D-014: API version | Locked |
Progress is not a claim made by the agent. Implementation, independent verification and release acceptance are tracked separately. Every result can point to its source, test or commit.
Part III / Session 20
The capstone.
20
Capstone Discussion & Architecture Review
Student project presentations, architecture critique, feasibility reviews and project approval. No new instructional content.
Student feedback
Help shape EAG V4.
This is a proposed syllabus. We are collecting student feedback before finalizing the depth, assignments and practical emphasis.
Your feedback is anonymous. We do not ask for your name or email, or save your IP address with your response. Please avoid including personal information.