An AI engineering roadmap: from machine learning to useful LLM applications

AI engineering starts before the model

An AI application depends on more than a model call. Engineers need to prepare data, establish a baseline, choose an evaluation method, connect the model to a workflow and measure the result. A course roadmap should make those dependencies visible.

The Fullstop AI Engineering outlines both begin with Python and programming, then build toward data analysis, machine learning and deployed applications.

A 3-month route covers the main stages

The 12-week, 300-hour outline groups the work into three parts: Python, data and mathematics; machine learning and deep learning; then LLM applications, retrieval and deployment.

Projects include a customer-churn model, an image classifier, semantic search, a document extractor and a document assistant with citations. The outline also covers model evaluation, prompt injection, privacy and a validated read-only tool.

A 6-month route adds more time between stages

The 24-week, 600-hour outline separates data analysis and mathematics from machine learning, then gives deep learning and language models a dedicated module. Later modules cover model APIs, embeddings, semantic search, RAG, tool calling, evaluation, security and operations.

Its capstone is a deployed document assistant, supported by an evaluation report, source code and demo. The curriculum calls for testing the assistant against more than 50 questions and measuring quality, latency and usage cost.

Look for evidence at each step

1. Check that a machine-learning project has a baseline, held-out evaluation and error analysis.
2. For a retrieval assistant, ask how sources, citations, unsupported questions and access rules are tested.
3. For deployment, ask how latency, cost, logs, model versions and regressions are reviewed.

Choose time and depth deliberately

Both outlines state 25 hours per week. The shorter one compresses the progression; the longer one gives separate blocks to data, mathematics, model evaluation, deep learning and AI application quality.

Published materials do not state prerequisites, instructor support, assessment policies or delivery format. Confirm those details before choosing a schedule.