Private study notes

ATLANTIS interview prep

Everything needed for the five instruments, plus the cross-cutting brief. Each chapter can be read or played aloud, with American English voices and adjustable speed.

1. 1EU Competition Law, Regulation 1/2003 and ECN+The project's subject matter. What the prohibitions are, what cartels actually do, how agencies detect them, and the seven procedural provisions that collide with computational enforcement.42 min22 sections
2. 2The AI ActRegulation 2024/1689, the fairness strand's main instrument, and the chapter where your own research becomes the strongest card you hold. Includes the Annex III gap for public enforcement authorities.44 min21 sections
3. 3The Digital Markets ActEx ante obligations on gatekeepers, and the clearest case where the law has already outrun manual enforcement. Lighter than the first two chapters, with four public corpora nobody has coded.26 min11 sections
4. 4The Digital Services ActGoanta's territory, and the EU's first attempt to make algorithmic auditing a legal obligation. Also the home of the single best research project available to you: a public corpus of several hundred million statements of reasons.28 min9 sections
5. 5The GDPRRegulation 2016/679, and the instrument that pulls hardest against collecting data to train a model. Wisman's field, and the home of SCHUFA — the case that decides whether a human reviewer is really deciding.38 min18 sections
6. 6Probability and Statistics, from the Ground UpWritten for someone relearning. Every term defined before use, the distribution-by-data-type table with plain-language meanings, the three tests your own paper runs with the arithmetic shown, and calibration as the bridge to a legal standard of proof.56 min18 sections
7. 7Machine Learning, from the Ground UpWritten for someone relearning rather than revising. Every term defined in plain words before it is used, with a worked example in procurement terms, and the audit question attached to each idea.58 min20 sections
8. 8NLP and Unstructured DataPreprocessing through embeddings, classification and sequence models, the evaluation metrics, and the part that matters most for this project: how you validate unstructured data, and why legal text is the hardest case there is.44 min15 sections
9. 9LLMs, RAG, RLHF and ReasoningHow a language model actually works, from tokens to decoding, then fine-tuning and RLHF, then retrieval-augmented generation and its failure modes. The chapter where your own paper is the content rather than a footnote.62 min17 sections
10. 10Logic and Symbolic ReasoningPropositional through first-order logic, deontic and defeasible logic, Prolog, argumentation frameworks and neurosymbolic methods. Your most differentiated asset, and the only approach that answers the duty to give reasons by construction.52 min15 sections
11. 11Deep LearningFrom a single neuron to transformers, with worked arithmetic. Includes GANs and graph neural networks, and why graphs are the architecture that actually matters for competition work.54 min16 sections
12. 12Frontier Topics & JargonThe words that get used in these conversations without ever being defined — agents, digital twins, mixture of experts, quantisation, sandboxes, model cards, drift. Each defined plainly, with what it means for enforcement and where the hype outruns the evidence.42 min12 sections
13. 13The Interview BriefTwenty-five minutes, four people, one opener. The question bank with answers, the story about the time something broke, the three questions to ask them, and the six traps to avoid.38 min9 sections