VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding
u_Unrolled

Build Atlas · the core course

One road. One project. Zero to AI engineer.

Nine parts, in order. Each one adds a working piece to Atlas, an AI assistant you build from scratch: from arrays of numbers to a GPT you train yourself, then fine-tuning, retrieval, tools and serving. 9 of 9 parts are ready to build, with 59 Library lessons behind them.

Code along

Every written part has a runnable checkpoint: a script, tests and the real numbers it produces on a laptop CPU. Part 4 trains a 1M-parameter GPT on the lessons themselves; Part 7 answers questions with citations.

Your progress building Atlas

0 of 9 parts built

Start herePart 1: Data as tensors
  1. 1

    Part 1: Data as tensors

    Everything later is arrays of numbers. Get fluent with them first.

    Build it · 5 steps · 10 min · freeTurn a folder of Markdown lessons into clean sections, question–answer pairs and batches of integers.
    Go deeper: 3 Library lessons teach the ideas0/3 read
  2. 2

    Part 2: Learning from data

    Before neural networks: what "learning" means, and how not to fool yourself measuring it.

    Build it · 5 steps · 10 min · freeA first model that learns: TF-IDF features, softmax regression in NumPy, and an evaluation that can't cheat.
    Go deeper: 3 Library lessons teach the ideas0/3 read
  3. 3

    Part 3: Neural networks from scratch

    Derive backprop, then let PyTorch do it for you, knowing exactly what it does.

    Build it · 5 steps · 10 min · ProBuild backpropagation yourself, prove it against finite differences, then write the training loop every later part reuses.
    Go deeper: 4 Library lessons teach the ideas0/4 read
  4. 4

    Part 4: From text to GPT

    Tokens, attention and next-token prediction, then the real thing, end to end.

    Build it · 4 steps · 10 min · ProTrain a byte-level BPE tokenizer and a 940K-parameter GPT on the lessons, on a laptop CPU, and read its loss curve honestly.
    Go deeper: 7 Library lessons teach the ideas0/7 read
  5. 5

    Part 5: Make it helpful

    A base model predicts text. Post-training turns it into an assistant.

    Build it · 4 steps · 10 min · ProTeach TinyGPT to answer questions with a chat template and a loss mask, then compare full fine-tuning with LoRA on unseen questions.
    Go deeper: 5 Library lessons teach the ideas0/5 read
  6. 6

    Part 6: Talk to it well

    Most of an AI product is what the model sees, and proof that it works.

    Build it · 5 steps · 10 min · ProPrompts as code with untrusted text escaped, one interface for every model, and an eval harness that names the cases a change breaks.
    Go deeper: 4 Library lessons teach the ideas0/4 read
  7. 7

    Part 7: Ground it in your documents

    Models forget and make things up. Retrieval gives them the facts, and evaluation keeps them honest.

    Build it · 5 steps · 10 min · ProChunk the lessons by structure, search them with BM25 and dense vectors, fuse the rankings, and measure retrieval apart from answers.
    Go deeper: 7 Library lessons teach the ideas0/7 read
  8. 8

    Part 8: Let it act

    From answering to doing: tool use, planning, and guardrails.

    Build it · 4 steps · 10 min · ProGive Atlas tools, memory and a loop, then put a gateway between the model and the tools so that a fooled model still can't do damage.
    Go deeper: 10 Library lessons teach the ideas0/10 read
  9. 9

    Part 9: Serve it fast

    Make it fast and affordable for thousands of users.

    Build it · 5 steps · 10 min · ProAdd a KV cache to TinyGPT, measure what batching buys, simulate static against continuous batching on real traffic, and turn throughput into a cost per token.
    Go deeper: 16 Library lessons teach the ideas0/16 read

Foundations

Need the basics first?

Dip into these whenever Build Atlas assumes something you haven't met yet.

Electives

Go deeper after a part

Each branches off Build Atlas at the part it builds on.