Recsys

3 Modern Approaches to Solving Cold Start in RecSys

Contextual bandits, meta-learning, and LLMs — how Spotify, TikTok, and YouTube handle new users and items. The practitioner's guide to cold start.

5 hrs ago • Rahul Agarwal

From Candidates to Clicks: The Engineering Anatomy of Ranking

How modern recommendation systems go from 1,000 candidates to the one item you actually tap

Mar 14 • Rahul Agarwal

Vector Search at Scale: The Production Engineer's Guide

IVF partitions space. PQ compresses memory. Together they make 100M vector search actually possible

Feb 3 • Rahul Agarwal

How YouTube Finds Your Next Video in Milliseconds

A deep dive into two-tower retrieval, in-batch negatives, and the tricks that make it work

Jan 26 • Rahul Agarwal

The 3-Stage Funnel Behind Every Modern Recommender System

Two-Tower models, vector databases, cross-encoders—and how they work together at scale

Jan 20 • Rahul Agarwal

How Recommendation Systems Learned to Think

Recsys Series Part 2: From collaborative filtering breakthroughs to generative AI agents that can chat about your preferences and explain their…

Oct 4, 2025 • Rahul Agarwal

RecSys Fundamentals: The Art and Science of Digital Matchmaking

Recsys Series Part 1: Master the three core approaches that power every modern recommendation engine

Sep 27, 2025 • Rahul Agarwal

#nojs-banner { position: fixed; bottom: 0; left: 0; padding: 16px 16px 16px 32px; width: 100%; box-sizing: border-box; background: red; color: white; font-family: -apple-system, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol"; font-size: 13px; line-height: 13px; } #nojs-banner a { color: inherit; text-decoration: underline; } This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts