Linear Thompson Sampling worked end to end on the contextual-bandits running example: where the Gaussian belief comes from, how one draw is made, the exact probability of each choice, why it is LinUCB with a random multiplier, the cold-start trap that hides inside its tuning, and a tested implementation.

Bandits made personal: learning which action fits which user, not just which wins on average. One decision fully worked by hand with LinUCB, the same decision replayed with Linear Thompson Sampling, and the care that running personalization demands.

Running a multi-armed bandit on real traffic: choosing between UCB and Thompson Sampling, rewards that arrive late, the inference adaptive allocation quietly breaks, and how to estimate regret when nobody knows the best arm.

Multi-armed bandits as adaptive decision-making: the explore-exploit trade-off, regret and its guarantees, one hand-worked Thompson sampling and UCB decision each, and a 200-seed simulation shoot-out.

What changes when clean random assignment is unavailable: the identification assumptions, estimands, and diagnostics behind Difference-in-Differences, Propensity Score Matching, and Instrumental Variables.

Retrieval-Augmented Generation (RAG) has rapidly become a widely used architecture for bridging the gap between static Large Language Models (LLMs) and dynamic, proprietary data. By fetching releva...

KV Cache: How LLMs Avoid Recomputing the Past Large language models generate text one token at a time. At every step, the model attends to all previous tokens. Naively, this would require recomput...

Anyone who has played with a Large Language Model (LLM) API has likely tweaked the temperature slider. The conventional wisdom is simple: “Low temperature = boring and factual, High temperature = c...

In the previous posts, we explored how an ad impression is auctioned in milliseconds and how machine learning models decide whether and how much to bid. However, every system we’ve discussed - from...

In the previous posts, we explored the plumbing of AdTech: how ad tags, redirects, pixels, and browsers coordinate to make a single impression possible. But for a data scientist or engineer, the mo...