How to Use This Map

What Praxia is — and is not

Praxia is an opinionated, curated route through the Data and AI landscape. It tells you what to learn, in what order, and from which sources — sequenced the way you would actually need to know things, not alphabetically or by how popular the topic was in 2023.

What it is not: a course. A bootcamp. A certification. A shortcut. The map describes a journey that takes most people three to five years of sustained, deliberate practice — from Foundations through AI Engineer — if they are working seriously alongside it. Someone who approaches it as a checklist to tick will reach the end of the list having learned the vocabulary but not the craft.

The research track is a branch, not a destination. Most people using this map are building toward practitioner roles — Data Analyst, Data Scientist, ML Engineer, AI Engineer. The research branch is for those who want to contribute new knowledge, not just apply existing knowledge. It requires everything the main track builds, plus depth in mathematics that most practitioners never need.

Navigating the map

Start at the overview map. It shows all five stages in sequence, with the research branch forking off Stage 2. Each node is a link. The left-rail route line (on desktop) shows where you are in the journey as you scroll — and you can mark stages complete to track progress.

Each stage page has nine sections, in a fixed order: what the role does, core concepts, mathematics required, tools and engineering, the project, resources, exit criteria, self-test questions, and a bridge to the next stage. Read the sections in order the first time. Return to individual sections as reference.

The mathematics curriculum lives at /mathematicsand is organised in three tiers. Tier 1 covers what you need for Foundations and early Stage 1. Tier 2 covers what you need for Stages 2 and 3. Tier 3 is research-grade depth. Stage pages link to specific sections — those links are not decorative. When a stage page says “understand this derivation,” follow the link.

Resources are never listed to be comprehensive. They are listed because they are the highest-ROI option for that stage, in a specific order. Read that order as a recommendation, not as items to eventually check off. The resources page lets you filter across all stages by type, cost, and stage.

The glossary defines every term the map uses. If you encounter an unfamiliar term on a stage page, look it up there first. Each glossary definition links back to the stage where the concept is taught in depth.

The study methodology

The method matters as much as the material. Watching a video course and feeling like you understood it is not learning. Learning is the capacity to do something you could not do before — under time pressure, without the worked example in front of you, in a context the tutorial did not cover. The following four practices are how that capacity builds.

Primary sources over tutorials

A tutorial explains a paper. A textbook explains a field. Neither replaces the paper or the field. Praxia lists tutorials and courses because they are valuable entry points — they give you the mental model to understand the primary source. But the ceiling of tutorial learning is the ceiling of the tutorial author’s understanding. Reading the primary source — the textbook, the original paper, the official documentation — removes that ceiling.

Concretely: if a stage page recommends a course alongside a textbook, the course is for intuition-building, the textbook is for substance. Do not finish the course and call the topic done. Return to the textbook. The chapters that were confusing during the course will have become accessible.

Spaced practice

Spaced repetition is not optional; it is how biological memory works. A concept encountered once, understood once, is mostly forgotten within a week. The same concept encountered again at increasing intervals — one day, three days, a week, a month — is consolidated. This is not pop psychology; it is well-replicated cognitive science.

In practice: use Anki or a similar tool for definitions and formulas you need to recall quickly. More importantly, return to earlier material as you progress. The data analyst who revisits probability distributions while studying machine learning finds them three times as comprehensible as they did the first time. The repetition is not a signal that you are slow — it is the mechanism.

Projects before you feel ready

The project sections on each stage page describe what to build. Build them — not because they are fun (some are not), but because project work is the only mechanism that converts declarative knowledge (“I know what a t-test is”) into procedural knowledge (“I can run a t-test correctly on data I have not seen before”).

Do not wait until you feel ready. You will never feel ready. The confusion that arises when you start a project is not a sign that you are not prepared; it is the confusion that produces learning. Sit with it. Debug it. The resolution is the lesson.

The projects are sequenced: each one requires skills from the previous. Do not skip them. Do not do them symbolically — a notebook that is half-finished, with a handful of cells and a README that says “in progress,” is not a portfolio piece and is not learning. Finish what you start.

Teaching to learn

Explain what you are learning — to a colleague, to a rubber duck, in a blog post, in a study group. The act of explanation forces precision: you cannot wave your hands when someone asks “wait, why does the gradient have to point downhill?” You either know or you discover that you do not.

The Feynman technique — explain a concept in plain language until you can do it without notes — is not a study hack. It is a diagnostic tool. The places where your explanation becomes vague are exactly the places where your understanding has gaps. Those gaps are what to go back and fix.

When to move on

Each stage page has an exit criteria section with a checklist and self-test questions. These are the gate, not the time estimate. The time estimates (3–6 months, 6–12 months) are honest averages for people who are working seriously — they will vary by prior background, available hours, and how hard the material hits you personally.

Do not move on because you are bored. Do not move on because you have covered all the material once. Move on when you can answer the self-test questions without notes and build the stage project without following a walkthrough. That is the bar. It is not arbitrary — it is calibrated to what the next stage will assume.

Do not stay because you are not confident. Confidence at this level of difficulty is a lagging indicator, not a leading one. If you meet the exit criteria, move on regardless of how uncertain you feel. The uncertainty is normal. It does not go away with more review of the previous stage; it goes away with exposure to the next one.

Honest framing

Praxia is not the fastest path to a job. If the goal is a job as quickly as possible, the fastest route is a focused bootcamp plus application practice plus networking — and there is nothing wrong with that goal.

This map is the path to genuine competence. The kind where you can sit across from a senior engineer and hold your own. Where you can identify the right statistical test without looking it up, debug a failing model without a tutorial, read a paper and understand what the authors actually did. That level of competence takes time — measured in years, not weeks.

The map can accelerate the journey by eliminating the wasted detours: the wrong resources, the wrong sequence, the rabbit holes that felt productive but led nowhere. It cannot eliminate the journey itself. Nobody learns this material by reading a map of it. They learn it by doing the work.