Visual roadmaps
Three routes through the same lessons. The quick path covers everything in order; the other two put either model training or AI applications first. Effort ranges are study estimates.
Quick path: complete grip roadmap
The full sequence in the shortest sensible order. Each step builds on the one before. Finish with a capstone that combines ML and AI.
-
Step 1 of 10: The big picture and your setup
4–6 h study estimateWhat AI and ML are, the project lifecycle, Python and .NET on your laptop.
- What AI, machine learning, deep learning and generative AI are Beginner 60 min
- How an AI or ML project runs from idea to production Beginner 75 min
- Set up your laptop for AI and ML in Python and .NET Beginner 1.5 h
-
Step 2 of 10: Data skills
8–12 h study estimateData types, collection, labelling, cleaning, exploring and features.
- Data basics: types, features, labels and splits Beginner 60 min
- Collecting and labelling data, with SQL Beginner 1.5 h
- Cleaning and exploring data Beginner 1.5 h
- Feature engineering: turning raw data into model inputs Intermediate 1.5 h
-
Step 3 of 10: Core machine learning
12–18 h study estimateHow models learn, regression, classification, the main algorithms and unsupervised learning.
- How a model learns: loss, training and overfitting Beginner 75 min
- Regression: predicting a number Beginner 75 min
- Classification: picking a category Beginner 75 min
- The main ML algorithms and when to use each Intermediate 1.5 h
- Unsupervised learning: groups, compression and anomalies Intermediate 75 min
-
Step 4 of 10: Measure, improve and apply ML
8–12 h study estimateMetrics, tuning, explanations, forecasting and recommenders.
- Measuring models honestly Intermediate 1.5 h
- Improving and explaining models Intermediate 1.5 h
- Time-series forecasting Intermediate 75 min
- Recommender systems Intermediate 75 min
-
Step 5 of 10: Deep learning
14–20 h study estimateNeural networks, training, vision, text, transformers, generation, speech, RL and graphs.
- Neural networks in plain words Intermediate 1.5 h
- Training deep networks well Intermediate 1.5 h
- Computer vision: classification, detection and segmentation Intermediate 1.5 h
- Text and sequences: tokens, embeddings and RNNs Intermediate 75 min
- Transformers and attention Intermediate 1.5 h
- Generative models: autoencoders, GANs and diffusion Advanced 75 min
- Speech and audio: recognition and text-to-speech Advanced 60 min
- Reinforcement learning Advanced 75 min
- Graph neural networks: learning from connections Advanced 60 min
-
Step 6 of 10: Large language models
14–20 h study estimateHow LLMs work, design, train on a laptop, scale up, fine-tune, run locally, evaluate.
- How large language models work Intermediate 75 min
- Tokens and sampling settings Intermediate 60 min
- Designing a language model Advanced 75 min
- Train a small language model on your laptop Advanced 2 h
- From laptop to big hardware: scaling training Expert 1.5 h
- Fine-tuning: SFT, LoRA, QLoRA and preference tuning Advanced 1.5 h
- Running language models on your own machine Intermediate 75 min
- Evaluating language models Advanced 60 min
-
Step 7 of 10: Talk to models and add knowledge
10–14 h study estimateLLM APIs, prompting, embeddings, RAG and advanced retrieval.
- Calling an LLM from your application Beginner 75 min
- Prompt engineering that works in production Beginner 60 min
- Embeddings and vector search Intermediate 75 min
- RAG: answering from your own documents Intermediate 2 h
- Advanced RAG: chunking, hybrid search, re-ranking and evaluation Advanced 1.5 h
-
Step 8 of 10: Agents, multimodal and trust
12–16 h study estimateTools, agents, frameworks, multimodal apps, evaluation, security and responsible AI.
- Tools and agents with human approval Advanced 2 h
- AI frameworks and MCP: what to use in Python and .NET Intermediate 60 min
- Multimodal AI apps: images, documents and voice Intermediate 60 min
- Evaluating AI applications Intermediate 75 min
- Security for AI applications Advanced 75 min
- Responsible AI: fairness, privacy, transparency and oversight Intermediate 60 min
- "Generative AI and Agentic AI: the whole picture" Advanced 60 min
-
Step 9 of 10: Production
10–14 h study estimateServing, MLOps, monitoring, cost and scale, architectures.
- Serving a model as an API Intermediate 1.5 h
- "MLOps: versioning, pipelines and safe releases" Advanced 1.5 h
- Monitoring models and AI apps in production Advanced 75 min
- Speed, cost and scaling Advanced 75 min
- Production architecture for AI systems Expert 2 h
-
Step 10 of 10: Capstones and real cases
30–60 h study estimateUse the decision guide on real cases and build at least two capstones.
ML engineer view roadmap
Focus on training models from data and running them in production. Read Part 6 later for LLM applications.
-
Step 1 of 9: Big picture and setup
4–6 h study estimateLifecycle and tools.
- What AI, machine learning, deep learning and generative AI are Beginner 60 min
- How an AI or ML project runs from idea to production Beginner 75 min
- Set up your laptop for AI and ML in Python and .NET Beginner 1.5 h
-
Step 2 of 9: Data
8–12 h study estimateCollect, clean, explore, engineer features.
- Data basics: types, features, labels and splits Beginner 60 min
- Collecting and labelling data, with SQL Beginner 1.5 h
- Cleaning and exploring data Beginner 1.5 h
- Feature engineering: turning raw data into model inputs Intermediate 1.5 h
-
Step 3 of 9: Supervised learning
8–12 h study estimateHow models learn, regression and classification.
- How a model learns: loss, training and overfitting Beginner 75 min
- Regression: predicting a number Beginner 75 min
- Classification: picking a category Beginner 75 min
-
Step 4 of 9: Algorithms and unsupervised learning
6–10 h study estimateWhich algorithm when; clustering, PCA and anomalies.
- The main ML algorithms and when to use each Intermediate 1.5 h
- Unsupervised learning: groups, compression and anomalies Intermediate 75 min
-
Step 5 of 9: Evaluation and improvement
6–10 h study estimateMetrics, validation, tuning, explanations.
- Measuring models honestly Intermediate 1.5 h
- Improving and explaining models Intermediate 1.5 h
-
Step 6 of 9: Forecasting and recommendations
6–10 h study estimateTwo of the most common business uses.
- Time-series forecasting Intermediate 75 min
- Recommender systems Intermediate 75 min
-
Step 7 of 9: Deep learning
14–20 h study estimateNetworks, training, vision and other model families.
- Neural networks in plain words Intermediate 1.5 h
- Training deep networks well Intermediate 1.5 h
- Computer vision: classification, detection and segmentation Intermediate 1.5 h
- Text and sequences: tokens, embeddings and RNNs Intermediate 75 min
- Transformers and attention Intermediate 1.5 h
- Reinforcement learning Advanced 75 min
-
Step 8 of 9: MLOps and production
10–14 h study estimateServe, track, monitor and scale models.
- Serving a model as an API Intermediate 1.5 h
- "MLOps: versioning, pipelines and safe releases" Advanced 1.5 h
- Monitoring models and AI apps in production Advanced 75 min
- Speed, cost and scaling Advanced 75 min
-
Step 9 of 9: Capstones
30–50 h study estimateChurn prediction, demand forecasting and image inspection.
AI engineer view roadmap
Focus on building products with language models: RAG, agents, evaluation and security. Learn enough ML to reason about models and data.
-
Step 1 of 9: Big picture and setup
4–6 h study estimateLifecycle and tools for Python and .NET.
- What AI, machine learning, deep learning and generative AI are Beginner 60 min
- How an AI or ML project runs from idea to production Beginner 75 min
- Set up your laptop for AI and ML in Python and .NET Beginner 1.5 h
-
Step 2 of 9: Enough data and ML
8–12 h study estimateData, how models learn, classification and evaluation.
- Data basics: types, features, labels and splits Beginner 60 min
- How a model learns: loss, training and overfitting Beginner 75 min
- Classification: picking a category Beginner 75 min
- Measuring models honestly Intermediate 1.5 h
-
Step 3 of 9: How LLMs work
6–10 h study estimateTransformers, tokens, sampling and limits.
- Transformers and attention Intermediate 1.5 h
- How large language models work Intermediate 75 min
- Tokens and sampling settings Intermediate 60 min
-
Step 4 of 9: Calling models and prompting
6–8 h study estimateAPIs, structured output, prompt patterns.
- Calling an LLM from your application Beginner 75 min
- Prompt engineering that works in production Beginner 60 min
-
Step 5 of 9: Embeddings and RAG
10–14 h study estimateSearch by meaning and answer from your documents.
- Embeddings and vector search Intermediate 75 min
- RAG: answering from your own documents Intermediate 2 h
- Advanced RAG: chunking, hybrid search, re-ranking and evaluation Advanced 1.5 h
-
Step 6 of 9: Agents and multimodal apps
10–14 h study estimateTools, agents, frameworks, images and speech.
- Tools and agents with human approval Advanced 2 h
- AI frameworks and MCP: what to use in Python and .NET Intermediate 60 min
- Multimodal AI apps: images, documents and voice Intermediate 60 min
-
Step 7 of 9: Quality, security and responsibility
8–12 h study estimateEvaluation, attacks and defences, fairness and privacy.
- Evaluating AI applications Intermediate 75 min
- Security for AI applications Advanced 75 min
- Responsible AI: fairness, privacy, transparency and oversight Intermediate 60 min
-
Step 8 of 9: Own and run models
10–16 h study estimateFine-tune, run locally, evaluate; serve and scale.
- Fine-tuning: SFT, LoRA, QLoRA and preference tuning Advanced 1.5 h
- Running language models on your own machine Intermediate 75 min
- Evaluating language models Advanced 60 min
- Serving a model as an API Intermediate 1.5 h
- Speed, cost and scaling Advanced 75 min
- Production architecture for AI systems Expert 2 h
-
Step 9 of 9: Capstones
30–50 h study estimateRAG assistant, approval-based agent and a local private assistant.