AI / ML Engineering
From data science basics to deploying LLMs in production
What you'll be able to do
- Train and evaluate machine-learning models
- Work with data using NumPy, pandas, and scikit-learn
- Build and fine-tune neural networks
- Deploy a model behind an API
Before you start
- Python fundamentals
- High-school level math (algebra); some statistics helps
- Curiosity about data and models
Level 1 ·Math & Python Foundations
Python for Data Science
NumPy, Pandas, Matplotlib: the data stack every ML engineer uses daily.
- Kaggle: Python Course (free)coursefree
- Kaggle: Pandas Course (free)coursefree
- Python Data Science Handbook (free online)docfree
- NumPy array operations & broadcasting
- Pandas groupby & merge
- Matplotlib + Seaborn EDA plots
- EDA on a real Kaggle dataset
Linear Algebra & Probability for ML
Vectors, matrices, dot products, probability distributions: the math underneath every model.
- 3Blue1Brown: Essence of Linear Algebra (YouTube)videofree
- Khan Academy: Statistics & Probabilitycoursefree
- fast.ai: Practical Deep Learning Part 2 (math)coursefree
- Matrix multiplication by hand & in NumPy
- Bayes theorem applied to a real problem
- Gradient descent visualised
Level 2 ·Classical Machine Learning
ML Fundamentals with Scikit-learn
Regression, classification, clustering, model evaluation, and feature engineering.
- Andrew Ng: Machine Learning Specialization (Coursera)coursefree
- Scikit-learn User Guidedocfree
- Kaggle: Intro to ML Coursecoursefree
- Train/val/test split strategy
- Cross-validation & hyperparameter tuning
- Feature importance & selection
- Titanic Kaggle competition (top 20%)
Feature Engineering & Data Pipelines
Imputation, encoding, scaling, and building reproducible sklearn pipelines.
- Kaggle: Feature Engineering Coursecoursefree
- Hands-On ML with Scikit-Learn (O'Reilly)coursepaid
- Pipeline with ColumnTransformer
- Target encoding vs. one-hot
- Outlier detection & treatment
Level 3 ·Deep Learning & LLMs
Deep Learning with PyTorch
Neural networks, CNNs, RNNs, training loops, and GPU acceleration.
- fast.ai: Practical Deep Learning for Coders (free)coursefree
- PyTorch Official Tutorialsdocfree
- Deep Learning Specialization: Andrew Ng (Coursera)coursefree
- Train a CNN on CIFAR-10
- Custom training loop with gradient clipping
- Transfer learning with ResNet
HuggingFace Transformers & Fine-tuning
Use pre-trained LLMs, fine-tune with LoRA/PEFT, and build text/embedding pipelines.
- HuggingFace NLP Course (free)coursefree
- HuggingFace PEFT Docs (LoRA)docfree
- Text classification with BERT
- Fine-tune with LoRA on a custom dataset
- Sentence embeddings with sentence-transformers
RAG Systems & LangChain
Retrieval-augmented generation: vector stores, embeddings, and multi-step chains.
- LangChain Python Docsdocfree
- Pinecone: Vector Database Docsdocfree
- DeepLearning.AI: LangChain for LLM App Development (free)coursefree
- Document ingestion + chunking pipeline
- Semantic search with pgvector
- RAG chatbot over custom docs
Level 4 ·MLOps & Deployment
FastAPI for ML APIs
Serve ML models as REST APIs with FastAPI, async loading, and background tasks.
- FastAPI Docsdocfree
- testdriven.io: FastAPI + MLarticlefree
- Model loading on startup (lifespan)
- Async prediction endpoint
- Response validation with Pydantic
MLflow, Modal & Model Deployment
Experiment tracking, model registry, and serverless GPU deployment with Modal.
- MLflow Docsdocfree
- Modal.com Docsdocfree
- HuggingFace Spaces (free GPU hosting)linkfree
- Log experiments + compare runs in MLflow UI
- Deploy model to Modal serverless GPU
- Gradio demo on HuggingFace Spaces
Frequently asked
Is the AI / ML Engineering roadmap free?+
Yes. The entire AI / ML Engineering roadmap and every curated resource is free to follow on Commit. You can track your progress, keep a daily streak, and earn a shareable certificate at no cost — there is no paywall.
How long does the AI / ML Engineering roadmap take to complete?+
About 180 hours of focused study across 9 courses and 4 stages. At roughly one hour a day that is about 6 months; you can move faster by studying more each day.
Do I get a certificate for finishing the AI / ML Engineering roadmap?+
Yes. When you complete the roadmap on Commit you receive a verifiable certificate of completion that you can add to LinkedIn and your public Commit profile as proof of what you finished.
Related roadmaps
Claude Code Mastery
Ship faster with agentic coding: workflows, MCP, and automation
View AI / ML EngineeringPrompting for Developers
Use AI to code, debug and ship faster without losing rigor
View Backend EngineeringBackend Engineering: Node.js
From zero to shipping production REST APIs
ViewMake it stick
Copy this roadmap into Commit and turn it into a tracked program with a streak graph, study logging, and a shareable certificate when you finish. Free forever.
Start AI / ML Engineering free