This roadmap is designed for absolute beginners with no prior experience in computer science or mathematics. It provides a step‑by‑step pathway through foundational topics into the most modern areas of artificial intelligence. Each section explains why the knowledge is important and lists free resources. Topics build on one another, so you can follow the tracks sequentially or dip into specific areas once you have the necessary prerequisites. To keep the journey approachable, jargon is explained in plain language and resources are marked with a difficulty tag:

• Easy – gentle introductions or beginner‑friendly materials.
• Medium – assumes some basic familiarity with the prerequisites.
• Hard – in‑depth or mathematically rigorous content.

The diagrams shown throughout (see /images folder) depict each track as a subway‑style path so that prerequisites and dependencies are easy to visualise. Feel free to print them or keep them nearby as you progress.

Each resource includes a difficulty tag and format indicator:

• Video – course, lecture series or recorded talk.
• Text – book, article or blog post.
• Interactive – coding exercises or labs you can try in your browser.

Tags are subjective and based on learner feedback. If you find a resource too advanced, revisit it later after completing earlier topics.

A strong foundation is the key to unlocking advanced AI topics. In this section you will build intuition about how computers work and how to reason mathematically about data and algorithms. Think of it as your north–south line on the subway map: everything else branches off of it. The diagram below summarises the core foundational subjects; each stop corresponds to a topic in the tables that follow. You can complete these in any order, but subjects like algebra and programming basics will make later topics easier.

These subjects underpin all modern computing. Take your time here; a strong grasp of these ideas will pay dividends later. By learning how machines represent and manipulate information, you will be prepared to optimise code, debug issues and design systems that scale. Each topic builds on the previous one: you start by writing simple programs, then understand how algorithms manage memory and time, and eventually explore how operating systems and networks coordinate complex tasks.

Mathematics provides the language of modern machine learning and data science. Build intuition before diving into heavy proofs. Many newcomers fear the math behind AI, but you do not need to be a mathematician to succeed. Instead, focus on the concepts: what does a derivative represent? How do matrices transform space? Why do probabilities add up to one? With intuition, you can later tackle more abstract work.

Data engineers build and maintain the pipelines that collect, process and store data. They ensure that raw logs and transactional tables are transformed into clean, structured datasets that analysts and models can use. This track begins with basic scripting and SQL, then gradually introduces the infrastructure used in industry: containers that package code, message brokers that handle streaming events, orchestrators that schedule jobs and warehouses that store terabytes of data. By the end, you will have built a mini data platform end‑to‑end.

• Python/Linux/SQL (Easy) – refresh your Python skills (see Programming Basics), learn command-line basics via The Missing Semester of Your CS Education, and practice SQL using SQLBolt or Mode Analytics SQL Tutorial.

• Containers & Cloud (Medium) – understand containers with the Docker Beginners Guide (official docs) and explore basic cloud storage using AWS S3 Getting Started or Google Cloud Storage Quickstarts.

• Ingestion & Streaming (Medium) – learn message queues and streaming via Apache Kafka 101 (Confluent free course) and practise building ingestion pipelines; explore cloud messaging services like Google Pub/Sub.

• Workflow Orchestration (Medium) – study orchestrators like Apache Airflow through the official tutorial and try the hands-on Prefect or Dagster tutorials.

• Warehouse & Analytics Engineering (Medium) – understand columnar warehouses and star schemas with Google BigQuery Fundamentals (Coursera audit) or Redshift Getting Started; learn dbt with the free dbt Fundamentals course and practise modelling tables.

• Batch & Monitoring (Medium) – delve into distributed computing via Apache Spark Free Intro (Databricks community edition) or Flink Getting Started (official docs); learn about data quality using Great Expectations tutorials and monitoring concepts from Monte Carlo blog posts.

• Capstone (Hard) – put it all together with Data Engineering Zoomcamp (DataTalks.Club) – a project-oriented nine-week course covering Docker, Terraform, GCP, Airflow, Kafka, Spark and BigQuery.

Business Intelligence focuses on turning data into actionable insights via dashboards and reports. Whereas data science asks “why does this happen?”, BI often asks “what happened and how can we measure it?”. Practitioners need strong communication skills, domain understanding and a sense of design to build dashboards that decision makers can trust. You will learn how to design schemas that support reporting, select appropriate charts, tell stories with numbers and empower non‑technical colleagues to explore data on their own.

• Spreadsheets & SQL (Easy) – master spreadsheet basics using Excel for Everyone (Coursera) (audit for free) or Microsoft’s Excel Training site. Practise formulas, pivot tables, and charts. Combine with SQL practice from the earlier fundamentals.

• Data Modeling (Medium) – learn to design star and snowflake schemas. Read overviews like this blog on Star Schema: The Complete Reference (Kimball methodology) and practise modelling exercises using sample datasets.

• BI Tools (Easy/Medium) – choose a tool and follow its official free training: Power BI Guided Learning, Tableau Free Training Videos, or Google Looker Studio Tutorials. Install the free versions and build sample dashboards.

• Dashboards & Storytelling (Medium) – watch Storytelling with Data Webinars and read the blog to learn about visual perception, chart design and narrative flow. Practise designing dashboards that highlight key metrics rather than clutter.

• Self‑Service BI (Medium) – explore open-source tools like Metabase or Apache Superset through their docs and community tutorials; learn how business users can explore data on their own.

• Advanced DAX & Reporting (Hard) – if you choose Power BI, deepen your knowledge of DAX using DAX Guide (SQLBI) and The Definitive Guide to DAX (YouTube series). Practise building complex measures and calculated tables.

Data scientists extract insights from data through exploration, analysis and modelling. They act as detectives: cleaning messy datasets, asking the right questions, applying statistical tests and building models to make predictions. This track emphasises statistics and practical data handling, guiding you through the entire lifecycle of a project—from loading CSV files to communicating findings. You’ll practice on open datasets, learn how to visualise patterns and understand the assumptions behind common models.

• Python (NumPy/Pandas) (Easy) – read the Python Data Science Handbook by Jake VanderPlas and work through exercises in Jupyter. Follow the official NumPy tutorials and Pandas tutorials.

• Data Wrangling (Medium) – practise cleaning messy datasets using the Pandas Cookbook on GitHub and Kaggle’s Data Cleaning micro-course; learn to handle missing values, outliers, and text formats.

• Exploration & Visualization (Medium) – learn exploratory data analysis using Kaggle’s Exploratory Data Analysis micro-course, the Matplotlib documentation, Seaborn documentation, and Plotly Express tutorials for interactive charts.

• Statistics & Probability (Medium) – revisit the probability & statistics resources above; supplement with StatQuest with Josh Starmer (YouTube) for intuitive explanations of statistical concepts.

• Hypothesis Testing & A/B Testing (Medium) – take the A/B Testing course from Google’s Data Analytics Professional Certificate (audit for free) or read blog posts explaining t-tests and p-values; practise designing experiments in a notebook.

• Time Series (Medium) – follow the book Forecasting: Principles and Practice (free online) and Kaggle’s Time Series micro-course. Learn ARIMA, exponential smoothing, and Prophet models.

• Feature Engineering & Causal Inference (Hard) – take Kaggle’s Feature Engineering micro-course; read Causal Inference for the Brave and True (free book) to understand confounding variables and potential outcomes; explore uplift modelling.

• Portfolio Project (Hard) – build a complete data science project by choosing an open dataset (e.g., from Kaggle, UCI Machine Learning Repository, or governmental sources) and walking through question formulation, cleaning, exploration, modelling, and communication.

Machine learning algorithms automatically learn patterns from data. They power everything from spam filters to recommendation engines. This track introduces supervised learning, where the goal is to predict labels, followed by unsupervised learning for discovering hidden structure, ensemble methods that combine many weak learners and deep learning for end‑to‑end function approximation. You will also learn how to evaluate models properly and how to deploy them into production with robust pipelines and monitoring.

• Supervised Learning (Easy/Medium) – take Machine Learning by Andrew Ng (Coursera audit) for an overview of linear regression, logistic regression, neural nets and SVMs; practise implementing these algorithms using scikit‑learn tutorials; watch StatQuest ML playlist for intuitive explanations.

• Evaluation & Metrics (Easy) – study confusion matrices, precision/recall, ROC curves and cross‑validation through scikit‑learn Evaluation Metrics documentation; practise measuring overfitting and underfitting.

• Unsupervised & Dimensionality Reduction (Medium) – learn clustering (k‑means, hierarchical), mixture models and dimensionality reduction (PCA, t‑SNE) via the Coursera Unsupervised Learning course (DeepLearning.AI) or Elements of Statistical Learning (free PDF) chapters; visualise high‑dimensional data using t‑SNE/UMAP tutorials.

• Advanced Models & Ensembles (Medium/Hard) – explore ensemble methods like Random Forests, Gradient Boosting, XGBoost and LightGBM through their open documentation and tutorials; watch StatQuest videos on decision trees and random forests.

• Deep Learning Basics (Medium) – take Neural Networks and Deep Learning (Coursera, part of the Deep Learning Specialization) or fast.ai Practical Deep Learning for Coders (free course) to understand backpropagation, activation functions and CNN/RNN basics; read Neural Networks and Deep Learning by Michael Nielsen (free book).

• Deployment & MLOps (Medium/Hard) – learn how to package and deploy models using TensorFlow Serving or TorchServe tutorials; explore model tracking with MLflow and orchestrate training pipelines with Kubeflow or Prefect; take Made With ML MLOps Course (free) or watch Full Stack Deep Learning lecture series.

Natural Language Processing (NLP) focuses on machines understanding and generating human language. It powers chatbots, translators and search engines. You will start with classic techniques like tokenisation and bag‑of‑words, progress to distributed representations such as word embeddings, and then explore neural sequence models. Modern NLP relies heavily on transformers and pre‑trained large language models; this track introduces these architectures and shows how to fine‑tune them for your own tasks.

• Text Processing & Tokenization (Easy) – read Natural Language Processing with Python (NLTK Book) to learn tokenization, stop-word removal and stemming; practise with NLTK and spaCy tutorials; watch CS224N Lecture 2 on text preprocessing.

• Word & Sentence Embeddings (Medium) – understand distributed representations via the Word2Vec Tutorial (TensorFlow blog) and GloVe Paper and Demo; learn sentence embeddings with SentenceTransformers (HuggingFace docs); read the Embeddings chapter in the Hugging Face Course.

• Sequence Models (Medium) – take Sequence Models by Andrew Ng (Coursera) to learn RNNs, GRUs and LSTMs; watch StatQuest: RNNs explained; implement simple language models using PyTorch or TensorFlow.

• Transformers & Attention (Medium/Hard) – read The Illustrated Transformer (Jay Alammar blog) for a gentle introduction to attention; follow the Hugging Face Transformers Course to build transformer models; watch lectures from Stanford CS224N on transformers.

• Transfer Learning & Fine-Tuning (Medium) – follow the Hugging Face Fine-Tuning Tutorial to adapt pre-trained BERT/GPT models to specific tasks; practise on text classification and question answering tasks; explore parameter-efficient methods like adapters and LoRA (Low-Rank Adaptation).

• Large Language Models & Prompting (Medium/Hard) – take the ChatGPT Prompt Engineering for Developers course (DeepLearning.AI) to learn best practices; read the OpenAI Cookbook for examples using the OpenAI API; explore LangChain tutorials to create chatbots that call tools and search engines.

Computer vision enables machines to interpret images and video. Cameras and sensors generate billions of pixels per second, and CV algorithms extract meaningful information from them. You will begin by learning how images are represented as matrices and how simple filters can enhance or detect edges. Then you will delve into convolutional neural networks for recognition and detection, eventually exploring cutting‑edge architectures like vision transformers. Applications range from medical imaging to self‑driving cars.

• Image Processing (Easy) – learn basics of image manipulation using the OpenCV‑Python Tutorials and Pillow; practise colour spaces, filtering and edge detection.

• Convolutional Neural Networks (CNNs) (Medium) – study CS231n: Convolutional Neural Networks for Visual Recognition (Stanford, Video & Notes) or the Deep Learning Specialization: Convolutional Neural Networks (Coursera, audit available); implement simple CNNs using PyTorch or TensorFlow.

• Object Detection & Segmentation (Medium/Hard) – follow the Computer Vision Specialization (DeepLearning.AI) for detection, localisation and segmentation; explore frameworks like YOLOv5, Detectron2, or Ultralytics Hub to build real‑time detectors; learn about semantic and instance segmentation with U‑Net and Mask R‑CNN.

• Transfer Learning & Pretrained Models (Medium) – read the PyTorch Transfer Learning Tutorial and use pre‑trained networks (ResNet, EfficientNet) as feature extractors; practise fine‑tuning on your own image datasets.

• Vision Transformers (Hard) – read the An Image Is Worth 16×16 Words paper and Jay Alammar’s Illustrated ViT; implement ViT using the Hugging Face Transformers library.

• Multimodal Vision (Hard) – explore models that combine vision and language such as CLIP (GitHub), DALL·E mini or Stable Diffusion; learn how image captioning and vision‑language pre‑training work.

Modern AI increasingly integrates multiple data types and leverages large language models with external tools. Multimodal models combine text, images, audio and video to produce rich understanding; retrieval‑augmented generation (RAG) supplements language models with external knowledge bases; autonomous agents use LLMs to decide which action to take next; and vLLM provides efficient inference for serving large models. This track introduces these cutting‑edge domains and offers practical tutorials so you can build your own chatbots, multimodal search engines and agents.

• Multimodal Basics & Fusion (Medium) – read the CMU Multimodal Machine Learning course notes to understand challenges like representation, alignment and fusion; watch lecture videos to see examples of audio‑visual models; explore Multimodal Machine Learning (MMML) tutorial slides and notes.

• Vision‑Language Models (Medium/Hard) – study CLIP via OpenAI’s blog post and code examples; learn about image captioning with Show and Tell or Show, Attend and Tell paper; experiment with HuggingFace implementations of CLIP, BLIP, LLaVA or Flamingo for vision–language tasks.

• Cross‑Modal Retrieval (Medium) – build systems that retrieve images from text or vice versa. Follow the HuggingFace cross‑modal retrieval tutorial and Weaviate’s Multimodal Search Guide. Understand how contrastive loss aligns embeddings across modalities.

• RAG & Vector Databases (Medium) – learn the concept of Retrieval Augmented Generation via LangChain’s RAG documentation and Build a RAG App. Experiment with vector databases like Pinecone, Weaviate, FAISS or Chroma using their free tiers and tutorials.

• Agents & Tool Use (Medium) – explore LangChain Agents tutorials to see how language models can decide which tool to call; study open‑source projects like BabyAGI and AutoGPT on GitHub to understand task decomposition and memory; practise building simple agents that search the web, perform calculations and summarise results.

• vLLM & High‑Performance Inference (Hard) – read the vLLM documentation for installation and examples; learn how paged attention and token streaming improve throughput; integrate vLLM with LangChain or FastAPI to serve models cost‑effectively. Compare with other inference engines like TensorRT-LLM or HuggingFace’s TGI.

This roadmap collates a wealth of free resources across the AI landscape. Remember to pace yourself, revisit challenging topics and build projects that interest you. As you progress, consider contributing to open‑source projects, joining online communities (such as Kaggle, HuggingFace forums or the DataTalks.Club Slack) and sharing your knowledge through blogs or talks. Continuous learning and curiosity are key to mastering modern AI.

Enjoy the journey!