Fourier transforms on graphs, spectral GCNs, and the heat equation: how diffusion operators connect graph signal processing to cold-start recommendation problems.

A/B testing in management consulting: frequentist and Bayesian methods, multiple testing corrections, and the specific challenges of testing recommender systems in a live production environment.

The mathematical essentials for recommender systems: from matrix factorization via SVD to graph convolutional networks, and why the spectral perspective unifies both approaches.

Bayesian methods in consulting practice: Thompson sampling for multi-armed bandits, conjugate priors for A/B testing, and hierarchical models for recommendation engines — with Python examples.