Waqas Bin Hamed

Currently building data pipelines and ML systems at The Home Depot. Computational math background from Waterloo — I care about how algorithms work, not just that they do.

What I’m currently working on, learning, and thinking about. Last updated May 2026.

Building

Spotify Playlist Organizer — An agentic LLM pipeline that classifies songs by mood and genre and automatically sorts them into Spotify playlists. Built with Gemini and LangChain, using structured output to keep categorizations consistent. Deployed on Google Cloud Run — still iterating on the classification logic. Turns out mood is harder to pin down than genre.

Exploring

Recommendation systems — How rec systems actually work at scale, and where they break. The standard approach optimizes on clicks or engagement, but that conflates correlation with causation — if someone clicks a song, did the recommendation cause it, or would they have found it anyway? That question leads naturally into causal inference.

Causal inference — Moving past “what correlates with what” to “what actually drives what.” Working through the theory and thinking about where it applies practically: evaluating recommendations, measuring the real effect of a pipeline change, understanding what a model is actually learning vs. what it’s picking up by accident.

AI-assisted workflows — Figuring out where tools like Claude Code genuinely save time vs. where they add noise. Less about generating code automatically and more about using them as a thinking partner for architecture decisions, debugging, and writing. Still calibrating this.

This page is inspired by the /now page movement.