A simulated case study inspired by Duck.ai: measuring a launch campaign with a Bayesian geo experiment on aggregate counts by region, designed in advance and graded against a known answer.
A Bayesian media mix model in NumPyro for five marketing channels on simulated regional data: whether the data can tell the channels apart, making the sampler work, and every channel's cost graded against the truth.
Calibrating a media mix model with a geo experiment by reweighting, without counting anything twice, then testing the budget it recommends against the truth and choosing the next experiment.
A case study on public Steam reviews: how long VRChat reviewers keep playing, the three decisions that make it measurable, and who stays.
A case study on public Steam reviews: the shape of the risk of leaving, a 2025 shift, what playtime, thumbs and language add, and a model checked on held out reviewers and later years.
A case study on public Steam reviews: turning retention patterns into a fair, well sized experiment for new players at the moment of a negative review.
Part 1: Introduction to Agents and Their Importance What Are Agents? Agents are autonomous systems built on top of large language models (LLMs) that can perform actions. Unlike LLMs, which respon...
When managing a project with tools like GitHub, Poetry, Ruff, and Black, having a streamlined workflow is essential. This guide will help you navigate the commands and best practices for version co...