Thirteen AI Courses, From First Principles to Agents
By glitchdata Team 2 min read
The course catalogue has been rebuilt from the ground up around AI. Every course has four short, written lessons, and the first is always a free preview.
Where to Start
If you're new to the field, begin with How machine learning works — the core ideas with no maths and no code — and Planning an AI project, a practical guide for anyone asked "can we use AI for this?". Prompting large language models covers what LLMs actually do and how to get reliable results from them.
Working With Data
Working with datasets for AI walks through reading a dataset card, loading CSV and Parquet files with pandas, handling missing values and checking licences — using real files from the hub. Forecasting time series uses NASA's temperature record and Capital Bikeshare rentals to cover baselines, honest time-based testing and gradient boosting.
Building Models
Your first classifier with scikit-learn trains, compares and evaluates a model on the Palmer penguins data. Neural networks and deep learning builds a network in PyTorch on the wine quality dataset. Computer vision with pretrained models runs the hub's ResNet-50 with ONNX Runtime, and Speech to text with Whisper transcribes audio and measures accuracy with word error rate.
Generative AI and Agents
Embeddings, semantic search and RAG builds retrieval from the ground up with all-MiniLM-L6-v2. Building AI agents with tool use explains the loop behind every agent framework, and Evaluating LLM applications shows how to test them properly. These three are deliberately provider-neutral, so they won't date when one vendor renames its SDK.
Why It Matters
Courses and data belong together. Learning to build a classifier is more useful on a real dataset you can open, inspect and cite — and every course here links straight to the hub entries it uses. Responsible AI in practice rounds out the set with fairness checks, privacy and model cards, because building AI well includes knowing where it can go wrong.