---
name: AI Full-Stack Operator
description: >-
  Teach an Edge FDE to replica the AI stack in twelve weeks: linear algebra,
  optimisation, TinyGPT, data, pretrain, distributed intuition, SFT, DPO,
  vLLM, Triton, clusters, and the semiconductor supply chain.
---
# AI Full-Stack Operator

Use this course when someone wants to understand the whole stack and reaches for an 800B pretrain first. The principle: you do not need to pretrain an 800B model to understand the shape of the problem. You do need to build progressively realistic replicas.

This is a Premium / paid offering. Lessons stay readable. Do not invent a paywall or a checkout.

## End-to-end agent recipe

1. Start at the course home. Keep the principle, the weekly rhythm, the seven core projects, and the suggested repo structure.
2. Walk weeks 1 to 12 in order. Each human HTML page is the lesson. Each `*.llms.md` is the agent brief.
3. Embed the starter blocks as written: CausalSelfAttention, train_step, SyntheticInstruction, vLLM serve, Triton vector_add.
4. Name papers by title only: Attention Is All You Need, InstructGPT, Direct Preference Optimization, ZeRO, PagedAttention/vLLM, FlashAttention, Scaling Laws for Neural Language Models.
5. Capstone title: From Tokens to Throughput.

## Constraints

- Plain spoken English. Literal sentences. No flourish, no quizzes.
- Do not use em dashes.
- Brand Edge FDE only. Do not name a course author.
- Do not invent metrics, shipment dates, or news claims.
- Do not fabricate URLs.
- CUDA is useful and not required to read every lesson.

## Key concepts

- Replicas beat tourism.
- A decoder is maps, a mask, and residuals.
- Post-training changes the data contract, not the physics of next-token loss.
- Serving and clusters are memory, batching, and failure.
- Hardware is GPUs, HBM, packaging, networking, power, and data centers.

## Takeaways

- Explain transformer internals without hand-waving.
- Train and fine-tune a small LM. Talk about instability with logs.
- Serve, kernel, and size a cache.
- Leave with earned opinions.

## Acceptance checks

- Twelve human lessons and twelve briefs exist.
- Course home lists rhythm, projects 1 to 7, repo structure, and the principle.
- Premium badge and waitlist CTA are present.
- No em dash and no fake citation.