What do you want to become?

Choose a goal and follow the courses in order. Hours are rough estimates; skip what you already know.

• 🧑‍💻

  AI Application Engineer

  Ship RAG and agent products with evals and guardrails

  About 90 hours

  1. 1🦙 Local LLM Inference — Core: foundations, runtimes, applications.
  2. 2🧪 LLM Evaluation and Testing — Golden datasets, judges, regression tests.
  3. 3🛡️ AI Security and Privacy — Threat model, injection, agent security.
  4. 4🚀 Serving AI in Production — Delivery track: CI, retries and FinOps.

  You can then: Build, evaluate and secure LLM features end to end.

• ⚡

  Inference / Performance Engineer

  Tune and operate high-throughput model serving

  About 170 hours

  1. 1🦙 Local LLM Inference — Runtimes and performance tuning.
  2. 2⚡ Inference Engineering — Bottlenecks, kernels, engines, techniques.
  3. 3🔧 Build an Inference Engine — Write an engine to understand one deeply.
  4. 4🚀 Serving AI in Production — Scaling and observability tracks.
  5. 5🎙️ Multimodal Inference — Audio, vision and embedding workloads.

  You can then: Predict, measure and improve tokens per second and latency.

• 🛠️

  LLM Platform / MLOps Engineer

  Run multi-tenant, monitored, autoscaled inference

  About 140 hours

  1. 1🦙 Local LLM Inference — Core: runtimes and serving.
  2. 2🚀 Serving AI in Production — All tracks: architecture, Kubernetes, scaling, observability, delivery.
  3. 3⚡ Inference Engineering — Engines, parallelism, disaggregation.
  4. 4🛡️ AI Security and Privacy — Stack and supply-chain security.
  5. 5🧪 LLM Evaluation and Testing — CI evaluation for releases.

  You can then: Operate GPU inference platforms with SLOs and cost control.

• 🧱

  ML Engineer / Model Builder

  Train, fine-tune, evaluate and ship models

  About 150 hours

  1. 1🧱 Build an LLM from Scratch — Tokenizer to trained transformer.
  2. 2🎛️ Fine-tuning and Customisation — LoRA, data, preference tuning, merging.
  3. 3🧪 LLM Evaluation and Testing — Measure what you built.
  4. 4🦙 Local LLM Inference — Quantisation track for deployment.

  You can then: Understand and adapt transformer models from the inside.

• 🧠

  Reasoning / Research Engineer

  Build and evaluate reasoning systems

  About 160 hours

  1. 1🧱 Build an LLM from Scratch — Prerequisite for the reasoning course.
  2. 2🧠 Build a Reasoning Model — Verifiers, voting, search, GRPO, distillation.
  3. 3🧪 LLM Evaluation and Testing — Honest evaluation.
  4. 4🎛️ Fine-tuning and Customisation — Distillation and preference tuning.

  You can then: Design verifiers, spend inference compute wisely, and understand RL for reasoning.

• 🛡️

  AI Security Engineer

  Threat-model, red-team and harden LLM systems

  About 110 hours

  1. 1🦙 Local LLM Inference — Core: how models are served and used.
  2. 2🛡️ AI Security and Privacy — All tracks, including red-team methodology.
  3. 3🧪 LLM Evaluation and Testing — Safety and adversarial evaluation.
  4. 4🚀 Serving AI in Production — Observability and incident response.

  You can then: Find and fix vulnerabilities in LLM applications and agents.

• 🦙

  Local AI Power User

  Run private assistants, voice and RAG at home

  About 60 hours

  1. 1🦙 Local LLM Inference — Hardware, quantisation, runtimes, apps.
  2. 2🎙️ Multimodal Inference — Vision and voice locally.

  You can then: A capable private AI setup on your own hardware.

• 🧭

  Tech Lead / Founder (survey)

  Make build-versus-buy, cost and risk decisions

  About 25 hours

  1. 1🦙 Local LLM Inference — Foundations and hardware sizing.
  2. 2⚡ Inference Engineering — Bottleneck maths lesson.
  3. 3🚀 Serving AI in Production — Capacity planning and FinOps lessons.
  4. 4🛡️ AI Security and Privacy — Threat model lesson.
  5. 5🧪 LLM Evaluation and Testing — Why vibes fail.

  You can then: Informed decisions about cost, capacity, quality and risk.