Course 4: MLOps for Scaling TinyML

← Back to TinyML Overview Download All (ZIP) edX Course

Chapter 5 · Course 4

MLOps for Scaling TinyML

Machine Learning Operations through the lens of TinyML: ML development, training operationalization, continuous training, model conversion, deployment at scale, prediction serving, continuous monitoring, and responsible AI.

Textbook companion: Vol I Ch 14–15

§5.1 Welcome to MLOps for Scaling TinyML

Topic	Type
What to Expect in This Course	Reading
Welcome Message	Slides
Who Should Take This Course?	Reading
The Past, Present and Future of ML	Slides
Why the Future of ML is Tiny and Bright	Reading
Machine Learning Lifecycle	Slides
Review of Course 1, 2 & 3	Reading
Scaling TinyML	Slides
Introduction to MLOps	Slides
Overview of MLOps	Reading
Course Structure	Slides
Course Activities	Reading
Your Mindset: T-Shaped Skills Needed for ML Engineers	Slides
Who's Who in MLOps for TinyML?	Reading

§5.2 MLOps: The Big Picture

Topic	Type
Overview of MLOps Objectives	Reading
What is MLOps, DevOps, and AI Ops	Slides
MLOps: A Use Case Overview	Slides
MLOps Persona	Reading
MLOps: Key Activities and Lifecycle	Slides

§5.3 ML Development

Topic	Type
Overview of ML Development	Reading
ML Development	Slides
Problem Definition	Slides
How Might You Define KWS	Reading
Data Selection for KWS	Slides
Why Real Data Matters	Reading
Data Exploration	Slides
Data Visualization Tools	Reading
Feature Engineering	Slides
Feature Engineering for KWS: A Case Study	Reading
Model Prototyping	Slides
Model Prototyping: Research vs. Production	Reading
Model Validation	Slides
Model Evaluation	Slides
Data Engineering	Slides
ML Development Impact on MLOps	Slides

§5.4 Training Operationalization

Topic	Type
Overview of Training Operationalization	Reading
Training Operationalization	Slides
CI/CD Triggers	Slides
Software Artifacts	Reading
Continuous Integration	Slides
CI Tools	Reading
Continuous Delivery	Slides
Production Deployment	Slides
Online Experimentation	Slides
Production Deployment in ML Deployment	Reading
Case Study Discussion	Reading
Training Operationalization Impact on MLOps	Slides

§5.5 Continuous Training

Topic	Type
Overview of Continuous Training	Reading
Continuous Training	Slides
Retraining Triggers	Slides
Data Processing Overview	Slides
Data Engineering for Everyone	Reading
Data Ingestion	Slides
Data Validation	Slides
Data Transformation	Slides
Training vs. Tuning	Reading
Training with AutoML	Slides
Neural Architecture Search (NAS) - Part 1	Slides
Neural Architecture Search (NAS) - Part 2	Slides
The Carbon Price of AutoML: CO2	Reading
Continuous Training with Transfer Learning	Slides
Pros and Cons of Transfer Learning	Reading
Continuous Training Metrics	Slides
Metrics for Continuous Training	Reading
Continuous Training Impact on MLOps	Slides
Optional: Multilingual Spoken Words Colab	Colab

§5.6 Model Conversion

Topic	Type
Overview of Model Conversion	Reading
Model Conversion	Slides
ML Frameworks & The Lay of the Land	Slides
TF vs. TFLite vs. TFLite Micro	Slides
TFLite Micro for TinyML	Reading
Model Pruning	Slides
Model Clustering	Slides
Model Quantization	Slides
Collaborative Optimizations	Reading
Student Teacher Networks / Knowledge Distillation	Slides
Model Conversion Impact on MLOps	Slides
Model Conversion Case Study - Smart DoorBell	Reading

§5.7 Model Deployment

Topic	Type
Overview of Model Deployment	Reading
Model Deployment	Slides
Scaling ML into Production Deployment	Slides
Containers for Scaling ML Deployment	Slides
Dockers vs. VMs	Reading
Challenges for Scaling TinyML Deployment (Part 1)	Slides
Challenges for Scaling TinyML Deployment (Part 2)	Slides
Challenges of Scaling TinyML Deployment	Reading
Anything As A Service	Reading
TinyMLaaS (Part 1): An Introduction	Slides
TinyMLaaS (Part 2): Design Overview	Slides
Summary of TinyMLaaS	Reading
Model Deployment Impact on MLOps	Slides
Driving Mode Detection Case Study	Reading

§5.8 Prediction Serving

Topic	Type
Overview of Prediction Serving	Reading
Prediction Serving	Slides
Prediction Serving Scenarios	Reading
Prediction Serving Scenarios: Batch	Slides
Prediction Serving Scenarios: Online	Slides
Prediction Serving Scenarios: Streaming	Slides
Prediction Serving Scenarios: Embedded	Slides
Prediction Serving Architectures	Slides
Embedded Inference Serving Benchmarks	Slides
Embedded Benchmarks: An Overview	Reading
MLPerf Tiny	Reading
Prediction Serving Impact on MLOps	Slides

§5.9 Continuous Monitoring

Topic	Type
Overview of Continuous Monitoring	Reading
Continuous Monitoring	Slides
Model Drift: The Big Picture	Slides
Concept Drift	Slides
Data Drift	Slides
Dealing With Drift	Slides
Continuous Evaluation Challenges for TinyML	Slides
TinyML Communication Challenges & Technologies for Continuous Monitoring	Reading
On-device Training: Limitations and Opportunities	Reading
Continuous Monitoring with Federated ML	Slides
Federated Learning Gboard	Reading
Continuous Monitoring Impact on MLOps	Slides
The Privacy vs Performance Trade Off	Reading
Optional: Federated Learning Colab	Colab

§5.10 Data & Model Management

Topic	Type
Model vs. Data Management	Reading
Data and Model Management	Slides

§5.11 Responsible AI: Transparency & Sustainability

Topic	Type
Responsible AI Overview	Reading
Sustainability of TinyML	Slides
Sustainable AI	Reading
Model Cards for Transparency	Slides
TinyML for Social Impact	Slides

§5.12 Summary

Topic	Type
Course Summary	Reading
Key Concepts of MLOps	Slides
What's Next?	Slides

Note

These materials were originally developed for the HarvardX Professional Certificate in Tiny Machine Learning on edX. See the original curriculum for the full item-by-item breakdown including forum prompts and quizzes not listed above.