AI / ML Engineer Roadmap
A step-by-step path from math foundations and classical ML to deep learning, LLMs & RAG, MLOps, and shipping models to production, 22 stages, in the order you should learn them.
- 1
Internet
Understand how requests actually travel from a client to your server, since most models eventually ship behind an API.
0% - 2
Math Foundations
The math every model, loss function, and optimizer is built on.
0% - 3
Python for ML
The language and core libraries almost every ML workflow is written in.
0% - 4
Version Control Systems
Track changes to your notebooks and code and collaborate with other engineers.
0% - 5
Repo Hosting Services
Host your Git repositories and collaborate via pull requests.
0%Pick only one to get started, GitHub is trending - 6
Data Analysis & Visualization
Understand your dataset before you ever train a model on it.
0%Pick a visualization library - 7
Classical Machine Learning
The algorithms that solve most real-world tabular problems without a neural net.
0%Pick a library - 8
Model Evaluation
Know whether a model is actually good, or just good on the data you gave it.
0% - 9
Deep Learning Fundamentals
The building blocks every neural network, from a simple MLP to a transformer, is made of.
0% - 10
Deep Learning Frameworks
The framework you'll actually build and train models in.
0%Pick one to start with - 11
CNNs & Computer Vision
Teach a model to see: classify, detect, and localize objects in images.
0%Pick a library - 12
Sequence Models & NLP
Teach a model to read: from RNNs to the attention mechanism behind every modern LLM.
0%Pick a library - 13
Large Language Models
Work with pretrained foundation models instead of training from scratch.
0%Pick a provider - 14
Vector Databases
Store and search embeddings at scale for retrieval and semantic search.
0%Pick one - 15
Model Deployment
Turn a trained model into something other software can actually call.
0%Pick a tool - 16
MLOps & Experiment Tracking
Keep every training run, dataset version, and metric reproducible.
0%Pick one - 17
Model Monitoring
Know when a model in production has quietly stopped working.
0% - 18
Containerization
Package a model with its dependencies so it runs the same everywhere.
0%Pick one - 19
Cloud ML Platforms
Where training and inference actually run at scale.
0%Pick one to start with - 20
CI/CD for ML
Automate retraining and redeployment instead of shipping models by hand.
0%Pick a tool - 21
AI Ethics & Responsible AI
A model that's accurate but unfair or unexplainable is still a liability.
0% - 22
ML System Design
Design the data and training infrastructure around a model, not just the model itself.
0%
Keep going!
You're 0% of the way through. Every checkbox you tick is saved automatically in your browser.
Browse Tutorials