1. Pick a learning path
Each roadmap is a staged plan: what to learn first, what to build at each stage to prove it, and links to the articles that teach each step. If you are not sure where to begin, start with the one closest to your job or the job you want.
AI Engineer
From your first API call and token costs to structured output, tool use, retrieval with access control, evaluation and production agents.
Open the roadmap →Backend Engineer
One language and HTTP, relational data and transactions, failure handling with idempotency and retries, observability, the cloud, and integrating AI features.
Open the roadmap →Data Engineer
SQL and data modelling, batch processing with Spark and columnar files, lakehouse table formats, streaming, data quality, and data for AI.
Open the roadmap →ML Systems Engineer
How GPUs execute and move memory, the anatomy of a training step, mixed precision, parallelism, and fast inference and serving.
Open the roadmap →2. Or browse by track
Already know what you need? Jump into a topic area. Each one opens a section page listing every article in it.
AI, LLMs and the maths behind them
How models work, run and fail, from attention arithmetic to safe deployment.
Agents and agent protocols
Building agents that call tools, talk to each other and handle payments.
GPUs and ML systems
Hardware, kernels, training and serving at scale.
Backend and distributed systems
Designing services that stay correct and fast when things fail.
Data platforms
Batch, streaming and the Hadoop ecosystem.
Real-time media
Video, audio and two-way streaming on the web.
3. Learn by doing
Reading is half of it. The labs run in your browser with nothing to install: change a parameter and watch what happens to a cache, a queue, a GPU kernel or an agent.
Interactive labs
Simulations and tools you can poke at, each with notes on what to try.
Open the labs →Agent skills
Ready-to-use skill definitions for coding agents, with what each one does.
Browse skills →Research digest
A short, sourced daily summary of what changed in AI and infrastructure.
Read the digest →4. How to get the most out of an article
- Read the intro and the diagram first. They give you the shape of the system before the details, so the rest of the article has somewhere to land.
- Work the example yourself. Every article has a worked example or code; redo it with your own numbers or run the code. That is where understanding turns into skill.
- Read the failure modes. Knowing how a design breaks is what separates someone who has read about a system from someone who can run it.
- Do the "what to do next" checklist. Each article ends with concrete next steps and links to the related articles to read after it.