## Podcast Platform — Publishing, distribution and audience analytics for independent creators.

- URL: https://www.naveentehrpariya.dev/projects/podcast-platform
- Industry: Media
- Client type: Startup
- Country: United States
- Role: Full Stack Product Engineer
- Duration: 6 months
- Status: Live
- Year: 2023

An end-to-end home for independent podcasters: upload an episode, get it transcoded, distributed to every major directory through RSS, and see who actually listened.

### Problem

Creators were stitching together a host, a transcription service and a spreadsheet. Nothing shared data, episode metadata was entered three times, and analytics stopped at raw download counts.

### Objectives

- Take an episode from upload to published without leaving the product
- Generate compliant RSS that every directory accepts on first submission
- Give creators listener analytics beyond a download number
- Keep storage and bandwidth costs predictable as libraries grow

### Solution

Uploads go straight to object storage with presigned URLs, so audio never passes through the API. A queued worker transcodes to streaming-friendly bitrates and writes back durations and waveforms. RSS is generated per show and cached at the CDN edge, and playback events stream into a rollup table that powers the analytics views.

### Features

- Direct-to-storage uploads with resumable transfer
- Background transcoding and waveform generation
- Spec-compliant RSS feeds cached at the edge
- Episode scheduling and drafts
- Listener analytics with retention curves
- Embeddable player for creator websites

### Stack

  - Frontend: Next.js, React, Tailwind CSS
  - Backend: Node.js, Express, BullMQ
  - Database: MongoDB, Redis
  - Cloud: AWS S3, CloudFront, Docker
  - Integrations: FFmpeg, OpenAI Whisper, Stripe

### Engineering challenges

### Large audio files through a small API

Problem: Routing hour-long uploads through the application server tied up memory and timed out on slow connections.

Approach: The client uploads directly to S3 with a presigned URL and only tells the API when the object exists, so the server handles metadata rather than bytes.

### Directories that reject imperfect feeds

Problem: Each podcast directory enforces slightly different RSS rules, and a rejected feed blocks distribution entirely.

Approach: Feed generation was moved behind a validator that runs the directory rule sets before a feed is ever served, so problems surface at publish time rather than at submission.

### Outcome

- Publishing collapsed from three tools into one flow
- Feed rejections caught at publish time instead of after submission
- Bandwidth costs became predictable by serving audio from the CDN rather than the origin

### What it taught

Keep heavy bytes off the application path. Once uploads and delivery moved to storage and CDN, the API stayed small enough to reason about.

Case study: https://www.naveentehrpariya.dev/projects/podcast-platform
More work: https://www.naveentehrpariya.dev/llms.txt