AI Infrastructure Engineer - Learning Path
🎓 Part of the free, open-source AI Career Curriculum ecosystem — Infrastructure · ML Engineering · AI Engineering · Governance · Startup · Workforce.
💜 Sponsor this curriculum — sponsorships keep the whole open-source AI Career Curriculum free and moving.
🎯 Overview
This repository contains a complete, production-ready learning path for becoming an AI Infrastructure Engineer. Through comprehensive modules, real-world projects, and production-grade code stubs with educational TODO comments, you’ll develop the skills needed to build, deploy, and maintain ML infrastructure at scale.
Repository Status: ✅ 100% COMPLETE - All modules and projects ready for learning!
What You’ll Master
- ✅ Build ML Infrastructure from scratch (Docker, Kubernetes, cloud platforms)
- ✅ Deploy Production ML Systems with auto-scaling and comprehensive monitoring
- ✅ Implement End-to-End MLOps pipelines (Airflow, MLflow, DVC)
- ✅ Deploy Cutting-Edge LLM Infrastructure (vLLM, RAG, vector databases)
- ✅ Scale Training with distributed systems and GPU clusters
- ✅ Monitor and Troubleshoot complex ML systems in production
- ✅ Optimize Costs across cloud providers (60-80% savings possible)
Why This Learning Path?
- 🎓 Industry-Aligned: Based on actual job requirements from FAANG and top tech companies
- 💻 Hands-On: Code stubs with TODO comments guide you through real implementations
- 🏗️ Production-Ready: Learn patterns used at Netflix, Uber, Airbnb, OpenAI
- 📊 Career-Focused: Directly maps to $120k-$180k AI Infrastructure Engineer roles
- 🚀 Progressive: 10 modules building from basics to advanced LLM infrastructure
- 🔥 Modern Stack: 2024-2025 technologies (vLLM, RAG, GPU optimization)
✨ What’s New
2026-05-27 — Layout standardisation:
- 🧹 Removed 10 empty root-level
mod-XXX-*/ placeholder directories. They were vestiges from a pre-refactor layout; all canonical module content has lived under lessons/mod-XXX-*/ for some time. The repo now matches the layout expected by the curriculum-runner audit (lessons/ for learning content, modules/ in the paired solutions repo).
- 🧹 Removed orphan
lessons/mod-101-foundations/exercises/solutions/ (a duplicate single-file index). Reference solutions live in the paired ai-infra-engineer-solutions repo; inline pointers throughout the lessons now link there directly.
May 2026 Update:
- 🧪 All 62 promised labs authored across all 10 modules (foundations → LLM infrastructure). Each lab is a substantive, runnable walkthrough with objectives, prerequisites, numbered steps, validation checklist, cleanup, and troubleshooting.
- 📒 Two new reading lists:
advanced-engineer-path.md and staff-engineer-path.md (9–18 months and 2–5 years respectively).
- 🧹 Structural cleanup: mod-101 lecture duplicates resolved, quiz placement consolidated, empty Makefile/pyproject populated with real content, CURRICULUM.md self-claim corrected to reflect actual completion state.
- 🔍 Honesty pass on CURRICULUM.md: the prior “100% Complete” claim has been replaced with a per-module exercise/lab accounting. Lectures and projects are excellent; exercises are 32 of 119 promised and being filled in over subsequent content drops.
Earlier:
- 📝 Comprehensive Quizzes for modules 102-110 (265+ questions)
- Module 102: Cloud Computing (mid-module + final, 50 questions)
- Module 103: Containerization (25 questions)
- Module 104: Kubernetes (30 questions)
- Module 105: Data Pipelines (25 questions)
- Module 106: MLOps (30 questions)
- Module 107: GPU Computing (25 questions)
- Module 108: Monitoring (25 questions)
- Module 109: IaC (25 questions)
- Module 110: LLM Infrastructure (30 questions)
- 📋 Technology Versions Guide - Complete specifications for 100+ tools
- 🗺️ Curriculum Cross-Reference - Mapping to Junior track
- 📈 Career Progression Guide - Engineer to Principal roadmap
📊 What’s Included
10 Complete Learning Modules (130 Files)
| Module |
Topic |
Hours |
Status |
Quiz |
| 01 |
Foundations |
50h |
✅ Complete (15 files) |
✅ 30Q |
| 02 |
Cloud Computing |
50h |
✅ Complete (11 files) |
✨ +50Q |
| 03 |
Containerization |
50h |
✅ Complete (14 files) |
✨ +25Q |
| 04 |
Kubernetes |
50h |
✅ Complete (13 files) |
✨ +30Q |
| 05 |
Data Pipelines |
50h |
✅ Complete (12 files) |
✨ +25Q |
| 06 |
MLOps |
50h |
✅ Complete (12 files) |
✨ +30Q |
| 07 |
GPU Computing |
50h |
✅ Complete (12 files) |
✨ +25Q |
| 08 |
Monitoring & Observability |
50h |
✅ Complete (11 files) |
✨ +25Q |
| 09 |
Infrastructure as Code |
50h |
✅ Complete (12 files) |
✨ +25Q |
| 10 |
LLM Infrastructure |
50h |
✅ Complete (12 files) |
✨ +30Q |
3 Production-Grade Projects (77 Files)
| Project |
Technologies |
Duration |
Files |
Status |
| 01: Basic Model Serving |
FastAPI + K8s + Monitoring |
30h |
~30 |
✅ Complete |
| 02: MLOps Pipeline |
Airflow + MLflow + DVC |
40h |
30 |
✅ Complete |
| 03: LLM Deployment |
vLLM + RAG + Vector DB |
50h |
47 |
✅ Complete |
Total Repository: 207 files | ~95,000+ lines of code | 500+ hours of learning content
🎓 Prerequisites
Option 1: Complete Junior Curriculum (RECOMMENDED)
If you’ve completed the Junior AI Infrastructure Engineer curriculum, you have ALL required prerequisites! ✅
The Junior curriculum covers:
- ✅ Python fundamentals & advanced concepts
- ✅ Linux/Unix command line mastery
- ✅ Git & version control workflows
- ✅ ML basics (PyTorch, TensorFlow)
- ✅ Docker & containerization
- ✅ Kubernetes introduction
- ✅ API development & databases
- ✅ Monitoring & cloud platforms
Duration: 440 hours (22 weeks part-time, 11 weeks full-time)
Option 2: Self-Assessment
Haven’t completed Junior curriculum? Use our comprehensive Prerequisites Guide to:
- Check your readiness with detailed skill checklists
- Identify knowledge gaps
- Get personalized learning recommendations
- Run automated skill assessment
Minimum Requirements
If self-studying, you must have:
- Python 3.9+ (intermediate level: OOP, async, testing, type hints)
- Linux/Unix CLI (bash scripting, processes, debugging)
- Git fundamentals (branching, merging, collaboration)
- ML basics (PyTorch/TensorFlow, training, inference, evaluation)
- Docker basics (images, containers, Compose)
- Kubernetes intro (pods, deployments, services)
👉 Not sure if you’re ready? Read the Prerequisites Guide for detailed assessment.
🚀 Getting Started
Quick Start
# 1. Clone repository
git clone https://github.com/ai-infra-curriculum/ai-infra-engineer-learning.git
cd ai-infra-engineer-learning
# 2. Create virtual environment
python3.11 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# 3. Install dependencies
pip install -r requirements.txt
# 4. Start with Module 01
cd lessons/mod-101-foundations
cat README.md
Learning Path
- Modules 01-02 (Foundations) - Start here if new to ML infrastructure
- Modules 03-04 (Core Infrastructure) - Docker and Kubernetes mastery
- Modules 05-06 (MLOps) - Data pipelines and ML operations
- Modules 07-08 (Advanced) - GPU computing and monitoring
- Modules 09-10 (Modern Stack) - IaC and LLM infrastructure
Detailed guide: GETTING_STARTED.md
📖 Curriculum Overview
Module 01: Foundations ✅
50 hours | 15 files
Build your foundation in ML infrastructure:
- ML infrastructure landscape and career paths
- Python environment setup and best practices
- ML frameworks (PyTorch, TensorFlow)
- Docker fundamentals and containerization
- REST API development with FastAPI
View Module 01 →
Module 02: Cloud Computing ✅
50 hours | 11 files
Master cloud platforms for ML:
- Cloud architecture for ML workloads
- AWS (EC2, S3, EKS, SageMaker)
- GCP (Compute Engine, GCS, GKE, Vertex AI)
- Azure (VMs, Blob Storage, AKS, Azure ML)
- Multi-cloud strategies and cost optimization (60-80% savings)
View Module 02 →
Module 03: Containerization ✅
50 hours | 14 files
Deep dive into containers:
- Docker architecture and best practices
- Multi-stage builds and optimization
- Docker Compose for multi-service applications
- Container registries and image management
- Security and vulnerability scanning
View Module 03 →
Module 04: Kubernetes ✅
50 hours | 13 files
Master Kubernetes for ML:
- Kubernetes architecture and components
- Deployments, Services, ConfigMaps, Secrets
- GPU resource management and scheduling
- Autoscaling (HPA, VPA, Cluster Autoscaler)
- Helm charts and GitOps with ArgoCD
View Module 04 →
Module 05: Data Pipelines ✅
50 hours | 12 files
Build robust data pipelines:
- Apache Airflow for workflow orchestration
- Data processing with Apache Spark
- Streaming data with Apache Kafka
- Data version control with DVC
- Data quality validation and monitoring
View Module 05 →
Module 06: MLOps ✅
50 hours | 12 files
Implement MLOps best practices:
- Experiment tracking with MLflow
- Model registry and versioning
- Feature stores and engineering
- CI/CD for ML models
- A/B testing and experimentation
- ML governance and best practices
View Module 06 →
Module 07: GPU Computing & Distributed Training ✅
50 hours | 12 files
Harness GPU power:
- CUDA programming fundamentals
- PyTorch GPU acceleration
- Distributed training (DDP, FSDP)
- Multi-GPU and multi-node training
- Model and pipeline parallelism
- GPU memory optimization
View Module 07 →
Module 08: Monitoring & Observability ✅
50 hours | 11 files
Build comprehensive observability:
- Prometheus and Grafana
- Metrics, logs, and traces (OpenTelemetry)
- Distributed tracing with Jaeger
- Alerting and incident response
- Model performance monitoring
- SLIs, SLOs, and SLAs
View Module 08 →
Module 09: Infrastructure as Code ✅
50 hours | 12 files
Automate infrastructure:
- Terraform fundamentals and best practices
- Pulumi for multi-language IaC
- CloudFormation for AWS
- State management and modules
- Multi-environment deployments
- GitOps workflows
View Module 09 →
Module 10: LLM Infrastructure ✅
50 hours | 12 files
Master cutting-edge LLM infrastructure (2024-2025):
- LLM serving with vLLM and TensorRT-LLM
- RAG (Retrieval-Augmented Generation)
- Vector databases (Pinecone, Weaviate, Milvus)
- Model quantization (FP16, INT8)
- GPU optimization for inference
- Cost tracking and optimization
View Module 10 →
🛠️ Projects
Project 01: Basic Model Serving System ✅
⭐ Beginner | 30 hours | ~30 files
Build a complete model serving system:
- FastAPI REST API for image classification
- Docker containerization with optimization
- Kubernetes deployment with monitoring
- Prometheus and Grafana dashboards
- CI/CD pipeline with GitHub Actions
Technologies: FastAPI, Docker, Kubernetes, PyTorch, Prometheus, Grafana
View Project 01 →
Project 02: End-to-End MLOps Pipeline ✅
⭐⭐ Intermediate | 40 hours | 30 files
Create a production MLOps pipeline:
- Apache Airflow DAGs (data, training, deployment)
- MLflow experiment tracking and model registry
- DVC for data versioning
- Automated model deployment to Kubernetes
- Comprehensive monitoring and alerting
- CI/CD with automated testing
Technologies: Airflow, MLflow, DVC, PostgreSQL, Redis, MinIO, Kubernetes
View Project 02 →
⭐⭐⭐ Advanced | 50 hours | 47 files
Deploy cutting-edge LLM infrastructure:
- vLLM/TensorRT-LLM for optimized serving
- RAG system with vector database (Pinecone/ChromaDB/Milvus)
- Document ingestion pipeline (PDF, TXT, web)
- FastAPI with Server-Sent Events streaming
- Kubernetes with GPU support
- Cost tracking and optimization
- Comprehensive monitoring
Technologies: vLLM, LangChain, Vector DBs, FastAPI, Kubernetes + GPU, Transformers
View Project 03 →
💰 Cost Considerations
Cloud Costs
All learning materials can be completed within free tier limits:
- AWS: 750 hours/month t2.micro + $300 credits (varies)
- GCP: $300 credit (90 days)
- Azure: $200 credit (30 days)
GPU costs (optional, for advanced projects):
- On-demand: $1-3/hour
- Spot instances: $0.30-1/hour (70% savings)
- Estimated total: $50-150 for complete curriculum
Optimization Tips
- Use spot instances for training (60-90% savings)
- Leverage free tiers across multiple cloud providers
- Delete resources when not in use
- Use local development where possible
📚 Resources
Included Documentation
- Comprehensive lesson materials with examples
- Code stubs with TODO comments for guided implementation
- Complete project specifications with architecture diagrams
- Quizzes and assessments for each module
- Best practices and design patterns
External Resources
Curriculum Documentation
🎯 Learning Outcomes & Career Impact
After Completion, You’ll Be Qualified For:
AI Infrastructure Engineer
- 💰 Salary: $120,000 - $180,000
- 🏢 Companies: Tech companies, AI startups, ML-focused organizations
- 📈 Demand: Very high (growing 35% year-over-year)
ML Platform Engineer
- 💰 Salary: $130,000 - $190,000
- 🏢 Companies: Large tech firms, enterprises with ML teams
- 📈 Demand: High (specialized role)
MLOps Engineer
- 💰 Salary: $110,000 - $170,000
- 🏢 Companies: All organizations doing ML at scale
- 📈 Demand: Very high (fastest growing ML role)
Skills You’ll Demonstrate
✅ Kubernetes expertise with GPU scheduling
✅ End-to-end MLOps pipeline implementation
✅ LLM infrastructure and RAG systems
✅ Distributed training and GPU optimization
✅ Production monitoring and observability
✅ Cloud platform mastery (AWS, GCP, Azure)
✅ Infrastructure as Code with Terraform
✅ Cost optimization strategies
📊 Repository Statistics
- Total Files: 207
- Estimated Lines: ~95,000+
- Modules: 10 (all complete)
- Projects: 3 (all complete)
- Learning Hours: 500+
- Technologies: 50+
Technology Stack Covered
Core Infrastructure:
Docker, Kubernetes, Terraform, Helm, ArgoCD
ML & Data:
PyTorch, TensorFlow, Apache Airflow, Apache Spark, Kafka, DVC
MLOps:
MLflow, Feature Stores, Model Registry, CI/CD
LLM Infrastructure:
vLLM, TensorRT-LLM, LangChain, Vector Databases (Pinecone, Milvus, ChromaDB)
Cloud Platforms:
AWS (EC2, S3, EKS, SageMaker), GCP (GCE, GCS, GKE, Vertex AI), Azure (VMs, AKS, Azure ML)
Monitoring:
Prometheus, Grafana, OpenTelemetry, Jaeger, ELK Stack
GPU Computing:
CUDA, NCCL, Multi-GPU training, Distributed training
🤝 Contributing
We welcome contributions! Please see CONTRIBUTING.md for:
- Bug reports and fixes
- Documentation improvements
- New exercises and examples
- Updated best practices
🆘 Getting Help
📜 License
This project is licensed under the MIT License - see LICENSE for details.
🌟 Success Metrics
Upon completion, you should be able to:
🚀 Next Steps After Completion
This curriculum prepares you for AI Infrastructure Engineer roles. For career progression:
-
Gain Experience (1-2 years)
- Work on production ML systems
- Handle incidents and on-call rotations
- Contribute to open-source ML infrastructure projects
-
Advance to Senior Engineer (2-3 years total)
-
Become an Architect (4-6 years total)
- Continue with the Architect track
- Design enterprise ML platforms
- Strategic technical leadership
Maintained by VeriSwarm.ai
评论