About the AI/ML Field Guide
What this site is, who it is for, how it is organised, and the rules we hold ourselves to.
The AI/ML Field Guide is an independent, platform-neutral guide to artificial intelligence and machine learning as it is practised by engineers. It is written for people who build and run models and for students who intend to. Its purpose is to be the place you open when you need to understand something properly: the mathematics that matters, the training methods, the models, the frameworks, the hardware, the evaluation, the deployment, and the industry around them.
It is the sister site of FPGA Guide, built on the same templates and the same rules. Where the two subjects meet, in hardware acceleration, the two sites link to each other.
Who it is for
Three groups, roughly.
- Engineers who already train or deploy models and want a reliable reference that is not tied to one framework’s documentation or one provider’s marketing.
- Software, data and embedded engineers who need to understand enough about machine learning to work alongside the people who do it, or to decide whether a model belongs in a product.
- Students and self-taught learners who want a structured path from the first classifier to fine-tuning and serving a language model, with code that runs.
We assume you can write a short program in some language and remember a little calculus and linear algebra. We do not assume you have trained a model before.
How it is organised
| Section | What it holds |
|---|---|
| Learn | Four tracks (beginner, intermediate, advanced, expert) of ordered lessons with diagrams, runnable Python and exercises |
| Topics | One hub per subject, from foundations to careers, that gathers lessons, reference pages, projects, news and papers on that subject |
| Platforms | One page per framework, model provider or hardware platform, plus the open-source ecosystem, with tooling, pricing, licensing and migration notes |
| Reference | Cheat sheets, a framework matrix, a glossary and a model database, each with a “last verified” date |
| News | Short briefs on model releases, framework versions and industry events, each linked to its sources and to a page that teaches the background |
| Research digest | A weekly summary of papers from arXiv, in plain language |
| Projects | Build-this-tonight projects that run end to end on a laptop in under 15 minutes |
| Events | Conferences, courses and community meetings, with dates and costs |
The 70/30 rule
At least 70 % of the words on this site, measured over a rolling 30-day window, are evergreen teaching or reference material. News is capped at the remaining 30 %. If a week is busy with model launches, we publish fewer briefs or merge them into a roundup rather than let news crowd out the teaching. Every news brief links to at least one evergreen page so that the news feeds the learning rather than replacing it.
Independence
The Field Guide takes no sponsorship from any lab, cloud or hardware company. At the time of writing there are no affiliate links and no advertising. When that changes, this page will say so before it happens, and sponsored content, if ever accepted, will be labelled on the page it appears on.
We name products as their makers write them and we say when a claim is a vendor’s rather than something we measured. Benchmark numbers from a launch post are reported as the vendor’s, with the benchmark named. Company and product names are trademarks of their respective owners and are used here only to identify those products. We have no relationship with OpenAI, Anthropic, Google, Meta, NVIDIA, Hugging Face or any other organisation mentioned, beyond being their users.
Where we criticise a tool, a model or a provider, it is because we found a sharp edge, and we try to say exactly where it is.
How it is made
The site is written and maintained by a small editorial team, referred to on these pages as “we” or “the editor”. Content is drafted with AI assistance, from sources that are read and summarised in new words, and every piece passes automated originality, fact-grounding and code-execution checks before a person reviews it. The full description, including what AI assistance means on a site about AI and where its limits are, is on the editorial and AI policy page.
Reference pages carry a “last verified” date. If you find something wrong, the corrections page explains how to tell us and what happens next. You can reach the editor through the contact page.