- Published on
AI Engineer tổng hợp
- Authors
- Name
- Trần Mạnh Thắng
- @TranManhThang96
AI Engineer tổng hợp
Giới thiệu
Series ghi lại quá trình học lộ trình 50 ngày từ Senior Software Engineer sang AI Engineer, tập trung vào GenAI, RAG, LLM application engineering, local LLM, MLOps và production deployment.
Phase 1: ML Foundation (Day 1-8)
- Day 1: AI Mindset cho Senior SE
- Day 2: Math đủ dùng cho ML
- Day 3: ML Fundamentals
- Day 4: Python ML Stack
- Day 5: Feature Engineering
- Day 6: Model Evaluation Metrics
- Day 7: Error Analysis, Data Leakage, Threshold Tuning
- Day 8: Mini-project - Customer Churn ML Pipeline
Phase 2: Deep Learning, NLP, Transformer (Day 9-16)
- Day 9: Neural Network từ Zero
- Day 10: PyTorch Fundamentals
- Day 11: Training Loop, Optimizer, Scheduler
- Day 12: NLP Fundamentals & Tokenizer
- Day 13: Attention Mechanism
- Day 14: Transformer Architecture
- Day 15: Hugging Face Ecosystem
- Day 16: Mini-project - Fine-tune PhoBERT/BERT Classifier
Phase 3: LLM Application Engineering (Day 17-24)
- Day 17: LLM Fundamentals
- Day 18: Prompt Engineering Thực Chiến
- Day 19: Structured Output & Function Calling
- Day 20: LLM App Architecture cho Production
- Day 21: Raw SDK vs LangChain vs LlamaIndex vs LangGraph
- Day 22: Agent Patterns với LangGraph
- Day 23: Security Basics Cho LLM App
- Day 24: Mini-project - AI Assistant có Tool Calling + Memory
Phase 4: Fine-tuning & Local LLM (Day 25-30)
- Day 25: Khi nào Fine-tune, khi nào dùng RAG
- Day 26: Dataset Preparation cho Instruction Tuning
- Day 27: LoRA/QLoRA Hands-on
- Day 28: Evaluation trước/sau Fine-tune
- Day 29: Local LLM - Ollama, llama.cpp, vLLM
- Day 30: Quantization & Deploy Local Model API
Phase 5: Production RAG (Day 31-40)
- Day 31: RAG Architecture
- Day 32: Embedding Models & Benchmark cho tiếng Việt
- Day 33: Vector DB Production
- Day 34: Chunking Strategies
- Day 35: Metadata, Citation, Permission-aware RAG
- Day 36: Hybrid Search Production
- Day 37: Reranking Cho Production RAG
- Day 38: Advanced RAG Patterns Production
- Day 39: RAG Evaluation Production
- Day 40: Mini-project - Production RAG System End-to-end
Phase 6: MLOps & Production AI (Day 41-47)
- Day 41: MLflow, Experiment Tracking Và Model Registry
- Day 42: Model Serving Với FastAPI, SSE Và Production Boundary
- Day 43: Docker/K8s/GPU Serving Cho AI Workload
- Day 44: Observability Cho LLM App
- Day 45: Cost Optimization Cho LLM/RAG Production
- Day 46: Guardrails
- Day 47: LLM Testing, Golden Set, CI/CD Cho Prompt/RAG