0% complete0 of 49 tasks done
🌱

Foundation Track

Understand AI systems, APIs, containers, and CI/CD before going deeper into infrastructure.

Phases 1-2 · Beginner
☸️

Platform Track

Build Kubernetes, IaC, GitOps, model lifecycle, and pipeline automation foundations.

Phases 3-4 · Intermediate
🛡️

LLMOps Architect Track

Master serving, RAG, MCP, evals, observability, security, and capstone architecture.

Phases 5-8 · Advanced

Interactive roadmap

Choose a phase and start building

01

6–8 days · Beginner

🧠 AI DevOps Foundations

Build the base mental model: LLMs, DevOps, APIs, Linux, Git, Python, and why AI systems need different operational controls.

0/6tasks complete

What you will achieve

  • Explain tokens, context windows, embeddings, inference, and RAG in simple language.
  • Use Git, Linux, Python, FastAPI, and basic API calls confidently.
  • Understand how AI DevOps differs from classic DevOps and MLOps.

LLM & GenAI basics

3 tasks
TokensContext WindowInferenceEmbeddingsRAGAgents

Engineering foundation

3 tasks
LinuxGitPythonFastAPIREST APIsSecrets
Milestone

You can explain how an LLM request flows from API/client to model response, including cost, latency, and security constraints.

02

7–10 days · Beginner

⚙️ Containers & CI/CD Automation

Learn how AI services are packaged, tested, scanned, built, and promoted through automated pipelines.

0/6tasks complete

What you will achieve

  • Containerize a Python AI service and run it locally.
  • Build GitHub Actions pipelines for tests, linting, image build, and vulnerability scan.
  • Publish versioned images to a registry.

Docker & image workflow

3 tasks
DockerfileComposeImage TagsContainer RegistryMulti-stage Build

CI/CD quality gates

3 tasks
GitHub ActionsQuality GatesTrivyGitleaksSBOM
Milestone

A code push can test, build, scan, and publish your AI service container automatically.

03

10–14 days · Intermediate

☸️ Kubernetes Platform Engineering

Move from local containers to production-style cloud-native deployment with Kubernetes, Helm, IaC, and GitOps.

0/6tasks complete

What you will achieve

  • Deploy AI services on Kubernetes using manifests and Helm charts.
  • Provision infrastructure using Terraform or OpenTofu.
  • Use Argo CD to implement GitOps deployment and rollback.

Kubernetes core

3 tasks
PodsDeploymentsServicesIngressSecretsGPU Scheduling

IaC + GitOps

3 tasks
TerraformOpenTofuHelmKustomizeArgo CDGitOps
Milestone

You can deploy, upgrade, rollback, and observe an AI microservice on Kubernetes using Git as the source of truth.

04

10–14 days · Intermediate

🔁 ML Lifecycle & Pipeline Automation

Control the full lifecycle of models, prompts, datasets, artifacts, experiments, and ML workflows.

0/6tasks complete

What you will achieve

  • Track experiments, artifacts, metrics, and model versions.
  • Use a model registry to promote models between environments.
  • Build repeatable ML or fine-tuning workflows using pipelines.

Model registry & artifacts

3 tasks
MLflowModel RegistryArtifactsLineagePromotion

Pipeline orchestration

3 tasks
Kubeflow PipelinesArgo WorkflowsDVClakeFSReproducibility
Milestone

You can trace a model from data and experiment run to registry version, artifact location, deployment, and rollback.

05

12–15 days · Advanced

🚀 LLM Serving & Optimization

Learn production inference: serving engines, batching, GPU memory, autoscaling, quantization, latency, throughput, and cost control.

0/6tasks complete

What you will achieve

  • Serve local and open-weight models with Ollama, llama.cpp, and vLLM.
  • Deploy model serving workloads using KServe or Kubernetes-native serving.
  • Benchmark latency, throughput, tokens/sec, time to first token, and GPU usage.

Local and edge inference

3 tasks
Ollamallama.cppGGUFQuantizationLocal Models

Production serving

3 tasks
vLLMKServeTritonRay ServeBentoMLAutoscaling
Milestone

You can deploy and benchmark a self-hosted LLM endpoint and explain the trade-offs between vLLM, KServe, Triton, Ollama, and llama.cpp.

06

10–14 days · Advanced

🔎 RAG, Vector DB & MCP Tooling

Build AI systems that connect to private knowledge, retrieval pipelines, databases, enterprise tools, and MCP servers.

0/6tasks complete

What you will achieve

  • Design a production-ready RAG pipeline with ingestion, chunking, embeddings, retrieval, reranking, and citations.
  • Operate vector databases with indexing, metadata filters, backup, restore, and access control.
  • Build and secure MCP servers for agent-tool connectivity.

RAG pipeline

3 tasks
ChunkingEmbeddingsVector SearchHybrid SearchReranking

MCP and tool connectors

3 tasks
MCPTool UseConnectorsAuthRate LimitsTool Governance
Milestone

You have a working RAG service connected to a vector DB and one secured MCP server exposed to an AI client.

07

12–15 days · Advanced

🛡️ Evaluation, Observability & Security

Turn AI systems from demos into reliable production systems using evals, telemetry, dashboards, alerts, and security controls.

0/6tasks complete

What you will achieve

  • Trace prompts, model calls, retrievals, tool calls, latency, cost, and failures.
  • Run automated LLM, RAG, and agent evaluations in CI/CD.
  • Harden AI systems against prompt injection, data leakage, insecure tool use, and excessive agency.

LLM evaluation and tracing

3 tasks
LangfuseDeepEvalRAGASpromptfooLLM Evals

Platform observability and security

3 tasks
OpenTelemetryPrometheusGrafanaAlertsOWASP LLM Top 10
Milestone

Your AI service has traces, dashboards, alerts, automated evals, security checks, and a rollback path.

08

15–20 days · Capstone

🏗️ End-to-End AI DevOps Capstone

Build portfolio-ready projects that prove you can operate AI systems end-to-end, not just run tools.

0/7tasks complete

What you will achieve

  • Ship a full AI DevOps platform using CI/CD, GitOps, model serving, RAG, evals, observability, and security.
  • Create public or internal documentation with architecture diagrams, runbooks, dashboards, and demo videos.
  • Demonstrate rollback, model promotion, quality gates, and incident response.

Portfolio projects

4 tasks
CapstonePortfolioRunbooksArchitectureDemo

Production readiness

3 tasks
RunbookSLORollbackThreat ModelCost Report
Milestone

You have 3–5 documented AI DevOps projects that show real deployment, monitoring, evaluation, security, and rollback.

Portfolio builder

Capstone projects for AI DevOps

Use these to show real applied skill, not just course completion.