The Problem

I follow a lot of different topics - music production, AI, MMA, news, startups, film. Keeping up meant visiting 10-15 news sites and YouTube channels daily. I wanted my news in podcast format, but existing options were either generic robot narration or hour-long debates I didn’t have time for.

So I built yourcast - a personalized AI podcast generator that creates custom 3-5 minute episodes tailored to your interests.

Features

yourcast features secure Google authentication, curated topic categories, and the ability to define your own custom topics. One click generates a personalized episode in under two minutes. The built-in player offers full playback controls, chapter timestamps, and source citations for every story - plus an archive to revisit your previous episodes.

How It Works

Data Pipeline

The system ingests 16,000+ articles daily from over 200 RSS feeds spanning technology, sports, entertainment, finance, and more.

To avoid redundant coverage, articles are deduplicated using semantic clustering with Google’s text-embedding-004. Each article gets vector-embedded, and stories with cosine similarity above 0.85 are grouped together. This ensures you get one comprehensive take on each story rather than five versions of the same news.

Multi-Agent Script Generation

The podcast scripts are generated using Google’s Agent Development Kit (ADK) with Gemini AI. Different agents handle:

• Story selection and prioritization based on your interests
• Script structure and narrative flow
• Tone and pacing adjustments
• Content summarization without losing key details

Audio Generation

Eight parallel text-to-speech workers process the script sections concurrently. This parallelization was crucial - it reduced generation time by over 7x compared to sequential processing. The final result is a fully playable episode with chapter markers and source links:

Infrastructure

The entire system runs serverless on Google Cloud:

• Cloud Run for the application containers
• Cloud SQL (PostgreSQL) for article storage and user data
• Cloud Tasks for managing the generation pipeline

This architecture auto-scales from zero to 1000+ instances based on demand, with podcast generation completing in under two minutes.

Key Learnings

Here’s what I took away from building this:

Parallelism is your best friend. Three places where parallelism saved me: agent execution with multiple TopicScriptAgents running simultaneously, TTS generation with 8 audio chunks processing at once, and RSS fetching with 10 parallel workers for feed discovery. Combined speedup: 3× faster generation.

Semantic deduplication is essential. News feeds have massive overlap. Without clustering similar stories, podcasts would repeat the same information from different sources.

Serverless is magic. I deployed 3 Cloud Run services, Cloud SQL, Cloud Tasks, and Cloud Scheduler. Total infrastructure management time? Zero hours.

This was my first time building a production-grade AI-powered application from end to end - from inception to shipping the final product. I can get into excruciating detail about the finer technical points (and I probably will), but first - why don’t you try it?

try yourcast!