Lecture library MLOps · 45 lecture hours

MLOps and AI Systems

A one-deck-per-chapter course on designing, building, deploying, operating, and governing machine learning and AI systems in production. Eight chapters, eight decks, and a project lab series.

8Chapters
45Lecture hours
9Slide decks
End-to-endAI systems

Course chapters

Showing 8 of 8 chapters

Chapter 015 hrs

AI System Architecture and ML Lifecycle

  • ML system objectives, requirements and design
  • ML lifecycle and production system architecture
  • Batch, online and real-time ML systems
  • Challenges, trade-offs and failure modes
ReadyOpen deck
Chapter 026 hrs

Data Engineering for ML Systems

  • Training data: collection, preparation and sampling
  • Data quality, distribution and validation
  • Feature engineering and feature management
  • Data versioning, pipelines and workflow orchestration
ReadyOpen deck
Chapter 036 hrs

Model Development, Evaluation and Management

  • Model development and training workflows
  • Experiment tracking and reproducibility
  • Model evaluation and optimization
  • Model versioning, registry and automated training
ReadyOpen deck
Chapter 046 hrs

Model Deployment and Serving

  • Model deployment and serving architectures
  • Batch, online and real-time inference
  • Containerized ML systems and infrastructure
  • Inference optimization, scaling and reliability
ReadyOpen deck
Chapter 056 hrs

MLOps and Automation

  • ML pipelines and workflow automation
  • Continuous integration, delivery and training
  • Testing and deployment automation
  • ML infrastructure and platform management
ReadyOpen deck
Chapter 065 hrs

Monitoring and Maintenance of ML Systems

  • ML system monitoring and observability
  • Data distribution shift and model drift
  • Model performance and production monitoring
  • Retraining, continual learning and system maintenance
ReadyOpen deck
Chapter 077 hrs

LLMOps and Agentic AI Systems

  • Production architecture of LLM applications
  • Prompt, model and evaluation management
  • Retrieval- and tool-augmented AI systems in production
  • Agentic workflows and tool use
  • LLM observability, cost and reliability
ReadyOpen deck
Chapter 084 hrs

Scalable and Responsible AI Systems

  • Scalability and distributed AI systems
  • AI security and privacy
  • Responsible AI and governance
ReadyOpen deck
No chapters match that search. Try terms like “architecture”, “feature”, “deployment”, “drift”, “RAG”, or “governance”.

Project labs — WatchTower

Eight labs, one running system. A CCTV surveillance platform grows from a local cloud in Lab 01 into a monitored, audited service in Lab 08 — in the same repository, with the same five commands.

One project

Requirements written in Lab 01 are still load-bearing in Lab 08. No reset between chapters.

Local AWS

Floci on :4566: S3, DynamoDB, SQS and CloudWatch with the same API calls as production.

Evidence, not screenshots

Every lab ends with make verify. A clean clone must reach a green result.

Open the project lab deck

Syllabus at a glance

Eight chapters, 45 scheduled lecture hours. Each row maps to one slide deck.

ChapterTopicHrsCoverage
01AI System Architecture and ML Lifecycle5ML system objectives, requirements and design · ML lifecycle and production system architecture · Batch, online and real-time ML systems · Challenges, trade-offs and failure modes
02Data Engineering for ML Systems6Training data: collection, preparation and sampling · Data quality, distribution and validation · Feature engineering and feature management · Data versioning, pipelines and workflow orchestration
03Model Development, Evaluation and Management6Model development and training workflows · Experiment tracking and reproducibility · Model evaluation and optimization · Model versioning, registry and automated training
04Model Deployment and Serving6Model deployment and serving architectures · Batch, online and real-time inference · Containerized ML systems and infrastructure · Inference optimization, scaling and reliability
05MLOps and Automation6ML pipelines and workflow automation · Continuous integration, delivery and training · Testing and deployment automation · ML infrastructure and platform management
06Monitoring and Maintenance of ML Systems5ML system monitoring and observability · Data distribution shift and model drift · Model performance and production monitoring · Retraining, continual learning and system maintenance
07LLMOps and Agentic AI Systems7Production architecture of LLM applications · Prompt, model and evaluation management · Retrieval- and tool-augmented AI systems in production · Agentic workflows and tool use · LLM observability, cost and reliability
08Scalable and Responsible AI Systems4Scalability and distributed AI systems · AI security and privacy · Responsible AI and governance

Reading

Core references for production ML/AI systems, MLOps, and LLM applications.

Chip Huyen — Designing Machine Learning Systems

O’Reilly, 2022. System design, data, deployment, and monitoring.

Textbook

Chip Huyen — AI Engineering

O’Reilly, 2025. Foundation models, RAG, agents, evaluation, and LLM systems.

Textbook

Google — MLOps and production ML best practices

Guides on testing, CI/CD/CT, monitoring, and production readiness.

Reference

Tooling docs: MLflow, Kubeflow, Feast, Evidently, LangChain/LlamaIndex, KServe

Practical references for tracking, pipelines, features, serving, monitoring, and LLM apps.

Reference

Using the decks

Every chapter is a self-contained reveal.js presentation served from the local vendored build in vendor/.

Navigate

→ / space next · ← previous · Esc overview · S speaker notes.

Present

Open any deck, press F for fullscreen. Notes open in a second window.

Export

Append ?print-pdf to a deck URL, then print to PDF (landscape, no margins).

Offline

Decks load local reveal.js, plugins, and KaTeX from vendor/.