JRE Recommender

Overview

Tagged and analyzed 2,000+ Joe Rogan Experience podcast episodes by guest, topics, and user interest — then built a Telegram bot and React web app to surface relevant episodes based on what the user cares about.

The Problem

With 2,000+ episodes spanning every topic imaginable, finding the right JRE episode is genuinely hard. No good discovery tool existed that matched guests and topics to actual user preferences.

My Role

Sole author — built the full pipeline from episode tagging and data cleaning to the recommendation logic, Telegram bot, and Flask-backed React web interface.

Key Features

• —Dataset of 2,000+ episodes tagged by guest, topic cluster, and content type
• —Recommendation engine matching user interests to relevant episodes
• —Telegram bot interface with 150+ active users/month
• —React + Flask web interface for browsing and filtering
• —Pandas-based data pipeline for episode processing and tag management

Challenges

Creating a tag taxonomy that was broad enough to be useful but specific enough to return relevant results. Too coarse and recommendations are random; too granular and nothing matches.

What I Learned

A good recommendation system is mostly a data quality problem. The algorithm matters less than how cleanly and consistently the underlying data is structured.