Every module in the curriculum, from Python foundations to production AI systems. Each module is independent — start where you are.
A welcome letter, your development environment, and a live portfolio site. Free for everyone — no account required.
The fundamentals every production engineer needs — Python internals, software design, data systems, and the math behind AI. Take any module independently or follow the full path.
Go from writing Python to understanding it — memory model, generators, type system, concurrency, and the patterns production code is built on.
Foundation · Module 2Understand NumPy and Pandas at the level that lets you write correct, fast, non-surprising numerical code — and read the output when something goes wrong.
Foundation · Module 3Package, test, containerise, and ship code the way engineers on production teams do it — Git, Docker, CI, structured logging, and secrets management.
Foundation · Module 4Build the API layer every AI system needs — HTTP, auth, async I/O, streaming, and the resilience patterns that keep services alive under load.
Foundation · Module 5Trace a user request from browser to AI model and back — and understand where every piece of data lives at each step: Postgres, Redis, object storage, and async queues.
Foundation · Module 6Design and query PostgreSQL at the level production systems require — schemas, indexes, query plans, transactions, and connection pooling.
Probability, linear algebra, and calculus — derived from first principles and implemented in NumPy. The foundation every ML engineer needs to reason about models, not just use them.
The probabilistic reasoning that underlies every ML model — distributions, Bayes, MLE, information theory, and hypothesis testing.
Math · Module 8See matrices as transformations, not grids of numbers. Build the geometric intuition that makes PCA, attention, and embeddings legible.
Math · Module 9Derive gradients, implement backpropagation from first principles, and understand the optimisation landscape well enough to debug training runs.
Build a real churn prediction pipeline from scratch — problem framing through production deployment, using a single dataset as the investigation arc across all four modules.
Before you touch a model, you need to frame the problem, audit the data, define success, and build a baseline worth beating.
Classical ML · Module 11Linear models, tree models, and ensembles — the algorithms that handle structured data in every production AI system alongside LLMs.
Classical ML · Module 12Clustering, dimensionality reduction, and anomaly detection — the techniques that find structure in data without labels.
Classical ML · Module 13Take a trained model from notebook to a versioned, monitored, containerised scoring service — and learn how classical ML and LLMs compose in real AI systems.
Neural networks, the transformer architecture, fine-tuning, inference, and evaluation — built from scratch and deployed to production.
Build, train, and debug feedforward networks — the foundation for understanding every architecture that follows.
Deep Learning · Module 15Implement every component a transformer is built from — attention, positional encoding, the transformer block, tokenisation, and pretraining — and understand the scaling laws that govern LLM behaviour.
Deep Learning · Module 16Adapt pretrained models to new tasks — full fine-tuning, parameter-efficient adaptation, instruction tuning, and the alignment techniques that make models follow instructions.
Deep Learning · Module 17Understand how LLMs generate tokens, why inference is expensive, and how to make it fast — caching, quantisation, speculative decoding, and memory-efficient serving.
Deep Learning · Module 18Measure what your model actually does — perplexity, benchmarks, task-specific evals, LLM-as-judge, and regression testing.
Build production LLM systems — from API fundamentals through RAG pipelines, agents, MCP, and evaluation — using the Anthropic API throughout.
Turn documents into searchable vectors — embedding models, vector databases, approximate nearest neighbour search, hybrid retrieval, and retrieval evaluation.
AI Engineering · Module 24Build a production RAG pipeline — query understanding, reranking, context assembly, citation, and a full evaluation harness.
AI Engineering · Module 25Build agents that plan, call tools, and orchestrate the full AI system stack — including classical ML scoring endpoints.
AI Engineering · Module 26Build the evaluation and observability stack for a full AI system — LLM evals, classical model monitoring, distributed tracing, and a drift dashboard.
AI Engineering · Module 27Design, ship, and operate AI systems that compose classical ML and LLMs — system design, prompt engineering at scale, cost optimisation, guardrails, and incident response.
Understand, run, fine-tune, and deploy open-weight models — from transformer internals through QLoRA training to vLLM production serving.
Read a model card and understand every field. Know the architecture variants, capability tiers, and licence constraints before choosing a model for production.
Open Source LLMs · Module 20Run, inspect, and benchmark open source models on real hardware — quantisation formats, prompting strategies, batching, and knowing when to move to a dedicated server.
Open Source LLMs · Module 21The end-to-end fine-tuning workflow on real hardware — dataset prep, QLoRA training, merging adapters, and evaluating what you actually got.
Open Source LLMs · Module 22Deploy open source LLMs as production services — serving configuration, structured output, prompt caching, multi-model routing, monitoring, and cost modelling.
DSA coding, ML coding, system design, and behavioral — leveled by engineer, senior, and staff. 629 problems across four modules.
19 patterns from sliding window to dynamic programming. Pattern-first, then problem banks at engineer, senior, and staff level.
Interview Prep · Module 214 ML topics from scratch in NumPy — implement gradient descent, k-means, attention, and more without a library.
Interview Prep · Module 3ML system design from feature stores to model serving. Frameworks for scoping, estimating, and defending a design under pressure.
Interview Prep · Module 4STAR stories, leveling signals, and the questions that separate senior from staff. Templates and worked examples at each seniority level.