AI App / Agent Architecture
Agent loops, workflow graphs, handoffs, memory, tools, multi-agent boundaries, and retrieval orchestration.
Bilingual knowledge system
Learn to design complete AI systems, not isolated model calls.
A course-style architecture guide synthesized from 17 AI repositories across agents, serving, training, RAG, evaluation, LLMOps, tools, and production governance.
17repositories
6architecture domains
12lessons
34bilingual deep dives
Architecture Map
The course connects application control, model runtime, data retrieval, training, evaluation, tools, security, and production operations into one decision model.
flowchart LR
User[Users and workflows] --> App[AI app and agent architecture]
App --> Tools[Tools, MCP, platform gateway]
App --> RAG[RAG and vector data]
App --> Serving[Model serving and inference]
Training[Training and adaptation] --> Serving
Serving --> LLMOps[Observability, evaluation, LLMOps]
RAG --> LLMOps
Tools --> Governance[Security and governance]
LLMOps --> Governance
Governance --> Release[Production readiness]
Six Domains
Each domain includes English and Vietnamese learning material plus detailed repository-grounded notes.
Model Serving / Inference
Runtime selection, batching, scheduling, quantization, token streaming, and serving operations.
Fine-Tuning / Training
Adapter tuning, distributed training, checkpoint recovery, optimizer state, and reproducibility.
RAG / Vector Database
Embeddings, chunks, metadata, indexing, tenancy, durability, hybrid retrieval, and search quality.
Observability / Evaluation / LLMOps
Tracing, scoring, feedback, datasets, lineage, prompt governance, and promotion gates.
Tooling / MCP / AI Platform
Tool contracts, MCP servers, provider gateways, self-hosted AI workspaces, and admin boundaries.
Learning Portal
Search the core course, toolkit, capstone, assessments, and deep dives from one place.
English course homepageStart the English learning path.
Trang khóa học tiếng ViệtĐường học đầy đủ bằng tiếng Việt.
12-lesson curriculumArchitecture lessons across six AI domains.
ProjectsHands-on architecture projects and deliverables.
Architecture toolkitCopy-ready ADRs, matrices, scorecards, and checklists.
Enterprise Knowledge CopilotConcrete end-to-end capstone scenario.
Assessment packExam, rubric, and answer key for architecture judgment.
Repository atlasCompare 17 repositories by architecture role and risk.
Toolkit, Capstone, Assessment
The repo is designed to be used, not only read. Copy the templates, run the capstone, and score the architecture.
Copy-ready templates
ADR, runtime matrix, RAG contract, LLMOps scorecard, security review, and production readiness checklist.
Enterprise Knowledge Copilot
A concrete architecture review copilot scenario that exercises all six domains in one product.
Architecture review exam
English and Vietnamese exam tracks with scoring rubric and answer keys for self-evaluation.
English Track
Use this path to study the full course in English.
Tiếng Việt
Đường học đầy đủ bằng tiếng Việt cho solution architect và senior developer.
Repository Deep Dives
The course is backed by bilingual architecture notes for the selected AI repositories.
| Domain | Repositories | Docs |
|---|---|---|
| AI App / Agent Architecture | OpenAI Agents Python, LangChain, AutoGen, LlamaIndex | Group 01 |
| Model Serving / Inference | vLLM, llama.cpp, Transformers | Group 02 |
| Fine-tuning / Training | PEFT, DeepSpeed | Group 03 |
| RAG / Vector Database | Qdrant, Chroma | Group 04 |
| Observability / Evaluation / LLMOps | Langfuse, Phoenix, MLflow, TruLens | Group 05 |
| Tooling / MCP / AI Platform | MCP servers, Open WebUI | Group 06 |